{
  "data": {
    "a": {
      "slug": "gemini-4-neon",
      "name": "Gemini 4 Neon",
      "vendor": "Google",
      "vendorUrl": "https://ai.google.dev",
      "company": "google.com",
      "kind": "model",
      "category": "inference",
      "summary": "Gemini 4 Neon is Google's middle Gemini 4 model, between Hydrogen and Carbon, for complex coding, knowledge work and science, on the paid tier of the Gemini API with a 1,048,576-token limit shared between input and output.",
      "url": "https://www.anchorterminal.com/tools/gemini-4-neon",
      "markdownUrl": "https://www.anchorterminal.com/tools/gemini-4-neon.md",
      "slimMarkdownUrl": "https://www.anchorterminal.com/tools/gemini-4-neon.min.md",
      "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/gemini-4-neon.json",
      "dns": "gemini-4-neon.grade.anchorterminal.com",
      "license": "Closed model, used through the Gemini API and Google's apps under Google's terms",
      "transports": [],
      "remoteUrl": "https://generativelanguage.googleapis.com/v1beta",
      "packages": [],
      "auth": "api-key",
      "authNotes": "An `x-goog-api-key` from AI Studio on a project with billing, since Neon is on the paid tier only, per Google's pricing page and What's new guide. Google AI Pro and Ultra plans can use it in AI Studio, and paid Pro and Ultra subscribers in Antigravity. On Google Cloud it's `gemini-4-neon` on the Agent Platform, with Cloud credentials.",
      "pricing": "usage",
      "pricingNotes": "$3 per million input tokens and $15 output, flat across the 1M-token window, with implicit caching taking cached input to $0.15, per Google's pricing page. Paid tier only, no free tier. Every processed token is billed, thinking and grounding instructions included (checked 2026-11-18).",
      "priceSummary": "Pay per use",
      "where": "hosted",
      "x402": {
        "level": "no",
        "evidence": "No machine payment. Paid use needs billing on a Cloud project, per the Gemini API listing (checked 2026-11-18).",
        "endpoints": []
      },
      "toolCount": null,
      "popularity": {
        "githubStars": null,
        "npmWeekly": null,
        "pypiWeekly": null,
        "asOf": "2026-11-18"
      },
      "docsUrl": "https://aistudio.google.com/docs/models/gemini-4-neon",
      "llmsTxt": "https://aistudio.google.com/llms.txt",
      "capabilities": [
        "inference.llm"
      ],
      "tags": [
        "official",
        "hosted",
        "model",
        "multimodal",
        "reasoning",
        "long-context",
        "tool-calling",
        "usage-priced",
        "card-required",
        "closed-source",
        "llms-txt",
        "new"
      ],
      "lastRelease": "2026-11-18",
      "graded": true,
      "anchor": {
        "graded": true,
        "score": 65,
        "grade": "B",
        "agentReady": false,
        "rank": 557,
        "ranked": true,
        "rankOf": 1559,
        "categoryRank": 25,
        "methodology": "1.0",
        "run": "2026-10-01",
        "scores": {
          "ergonomics": 93,
          "maintenance": 83,
          "payments": 30,
          "reliability": 45,
          "schema": 85,
          "security": 55,
          "transparency": 73
        },
        "assessment": {
          "confidence": "medium",
          "date": "2026-11-18"
        },
        "negative": 0,
        "verdict": "Google's middle Gemini 4 model, between Hydrogen and Carbon, at $3 and $15 per million tokens with a 1M-token shared budget, four thinking levels, function calling, structured output and implicit caching at 95% off. Paid tier only, sampling parameters are ignored, explicit caching isn't supported, no retirement date is stated, and every benchmark is Google's.",
        "bestFor": "Agents doing complex coding, long tool-heavy sessions and document-heavy knowledge work on the paid Gemini API, where Carbon is slower than needed and Hydrogen falls short.",
        "strengths": [
          "$3 input and $15 output per million tokens, flat across the 1M-token window, with cached input at $0.15",
          "1,048,576 tokens shared between input and output, with long decode to the full budget",
          "Four thinking levels, low to xhigh, with high the default",
          "Function calling, structured output, code execution, computer use, file search, URL context and Search and Maps grounding",
          "The same request schema as Carbon, Argon and Hydrogen, so switching is a model ID"
        ],
        "weaknesses": [
          "Paid tier only, so an agent can't start on a free key",
          "temperature, top_p and top_k are ignored, and out-of-range values return a 400",
          "No explicit context caching, and strict thought-signature rules on parallel calls",
          "No retirement date or minimum notice stated on the Gemini API or Google Cloud",
          "Benchmark figures are Google's own, and the full safety report is marked coming soon"
        ],
        "agentNotes": [
          "Call generateContent at https://generativelanguage.googleapis.com/v1beta, or the Interactions API, with model gemini-4-neon and an x-goog-api-key on a billed project",
          "Leave thinking_level at high and try medium for interactive work; reserve xhigh for hard multi-step problems",
          "Don't send temperature, top_p or top_k, and declare tools natively rather than asking for raw-text JSON calls",
          "Return every thought_signature unchanged, and never copy one across parallel functionCall parts",
          "Set background to true with max_output_tokens above 65,536 for long outputs, and allow HTTP timeouts of 15 minutes"
        ],
        "metrics": {
          "kind": "remote",
          "measured": false
        },
        "reviewCount": 0,
        "avgRating": 0,
        "history": [
          {
            "basis": "public evidence",
            "confidence": "medium",
            "grade": "B",
            "methodology": "1.0",
            "run": "2026-10-01",
            "runLabel": "October 2026 research run",
            "score": 65
          }
        ],
        "editorialScores": {
          "ergonomics": 93,
          "maintenance": 83,
          "payments": 30,
          "reliability": 45,
          "schema": 85,
          "security": 55,
          "transparency": 55
        },
        "provenanceScore": 91,
        "checked": {
          "categories": {
            "ergonomics": "2026-11-18",
            "maintenance": "2026-11-18",
            "payments": "2026-11-18",
            "reliability": "2026-11-18",
            "schema": "2026-11-18",
            "security": "2026-11-18",
            "transparency": "2026-11-18"
          },
          "last": "2026-11-18",
          "oldest": "2026-11-18",
          "stale": false,
          "staleAfterDays": 90
        }
      },
      "connect": {
        "http": "curl \"https://generativelanguage.googleapis.com/v1beta/models/gemini-4-neon:generateContent\" \\\n  -H \"x-goog-api-key: $GEMINI_API_KEY\" -H \"content-type: application/json\" \\\n  -d '{\"contents\":[{\"parts\":[{\"text\":\"hello\"}]}]}'"
      },
      "letme": {
        "capability": "https://letme.dev/inference.llm",
        "tool": "https://letme.dev/gemini-4-neon"
      },
      "sameCompany": [
        "gemini-4-argon",
        "gemini-4-carbon",
        "gemini-4-hydrogen",
        "vertex-ai",
        "embeddinggemma-2",
        "embeddinggemma-300m",
        "gemini-embedding",
        "vertex-ai-tuning",
        "cloud-run-gpus",
        "google-model-armor",
        "shieldgemma",
        "google-imagen",
        "gemini-omni-1-2-flash",
        "google-veo",
        "google-lyria",
        "google-speech-to-text",
        "google-cloud-tts",
        "gemini-live",
        "google-adk",
        "google-memory-bank",
        "google-secret-manager",
        "google-cloud-document-ai",
        "google-weather-api",
        "chrome-devtools-mcp",
        "puppeteer",
        "mcp-toolbox-databases",
        "google-maps-platform",
        "google-cloud-translation",
        "google-calendar-api",
        "firebase-cloud-messaging",
        "google-cloud-storage",
        "google-drive-api",
        "gemini-enterprise",
        "gemini-decision-1",
        "antigravity-cli",
        "gemini-cli",
        "harness2",
        "jules",
        "gemini-spark",
        "google-search-console",
        "google-ads-api",
        "google-forms",
        "google-sheets-api",
        "gmail-api",
        "google-people-api",
        "diffusiongemma-26b-a4b-it",
        "gemma-4-12b-it",
        "gemma-4-26b-a4b-it",
        "gemma-4-31b-it",
        "gemma-4-e2b-it",
        "gemma-4-e4b-it",
        "virustotal"
      ],
      "sameModel": [
        "gemini-api"
      ],
      "area": "models",
      "unitPrices": [
        {
          "item": "Gemini 4 Neon input",
          "unit": "1m-tokens",
          "usd": 3,
          "note": "Flat across the 1M window"
        },
        {
          "item": "Gemini 4 Neon cached input",
          "unit": "1m-tokens",
          "usd": 0.15,
          "note": "Implicit caching, 95% off input"
        },
        {
          "item": "Gemini 4 Neon output",
          "unit": "1m-tokens",
          "usd": 15,
          "note": "Thinking tokens included"
        }
      ],
      "provenance": {
        "legalEntity": "Google LLC",
        "domain": "google.com",
        "domainRegistered": "1997-09-15",
        "domainNote": "The endpoint is on googleapis.com, Google's API domain. google.com was registered in 1997.",
        "endpointOnVendorDomain": true,
        "terms": "https://aistudio.google.com/docs/terms",
        "privacy": "https://policies.google.com/privacy",
        "statusPage": "https://aistudio.google.com/status",
        "changelog": "https://aistudio.google.com/docs/changelog",
        "securityTxt": "valid",
        "checked": "2026-11-18",
        "score": 91
      },
      "pageJsonUrl": "https://www.anchorterminal.com/tools/gemini-4-neon.json",
      "live": {
        "slug": "gemini-4-neon",
        "probe": {
          "target": "https://generativelanguage.googleapis.com/v1beta",
          "method": "get",
          "lastAt": "2026-11-19T01:54:50.544738903Z",
          "lastOk": true,
          "lastStatus": 404,
          "lastMs": 36,
          "authRequired": false,
          "uptime24h": 100,
          "uptime30d": 100,
          "p50ms24h": 32,
          "p95ms24h": 130,
          "samples24h": 15,
          "samples30d": 15,
          "days": [
            {
              "date": "2026-11-19",
              "probes": 15,
              "ok": 15
            }
          ],
          "since": "2026-11-19T00:31:47.878022337Z"
        },
        "updatedAt": "2026-11-19T01:54:50.544738903Z"
      },
      "streak": {
        "days": 0,
        "from": "2026-11-19",
        "sinceWatching": true,
        "best": 0
      }
    },
    "answer": "Gemini 4 Neon scores 65 (B) on agent readiness against Jais 3's 55.9 (C), and leads in 5 of 7 scored categories. Jais 3 leads on payments \u0026 pricing.",
    "b": {
      "slug": "jais-3",
      "name": "Jais 3",
      "vendor": "Institute of Foundation Models (MBZUAI) and Inception42",
      "vendorUrl": "https://ifm.ai/blog/jais3/",
      "company": "ifm.ai",
      "kind": "model",
      "category": "inference",
      "summary": "Jais 3 is an open-weight family of Arabic and English reasoning models from MBZUAI's Institute of Foundation Models and Inception42, from 0.9B to 375B parameters with up to 512K tokens of context, released under Apache-2.0 for self-hosting.",
      "url": "https://www.anchorterminal.com/tools/jais-3",
      "markdownUrl": "https://www.anchorterminal.com/tools/jais-3.md",
      "slimMarkdownUrl": "https://www.anchorterminal.com/tools/jais-3.min.md",
      "jsonUrl": "https://www.anchorterminal.com/api/v1/tools/jais-3.json",
      "dns": "jais-3.grade.anchorterminal.com",
      "repo": "https://huggingface.co/inception42/Jais-3-375B-A23B",
      "license": "Apache-2.0, declared in the metadata of every model card read. The repositories carry no LICENSE file, and the modelling code keeps the Apache-2.0 header of the Qwen3-MoE code it was generated from",
      "transports": [
        "http"
      ],
      "packages": [],
      "auth": "none",
      "authNotes": "No account or key. The inception42 and IFM repositories on Hugging Face are not gated and download anonymously. A local server has whatever authentication the owner adds, such as vLLM's `--api-key`. The cards' examples call the local server with `api_key=\"EMPTY\"`.",
      "pricing": "free",
      "pricingNotes": "Free to download and run under Apache-2.0, with the owner's GPUs as the cost. No hosted endpoint from IFM or Inception42 was found, and on 4 November 2026 Hugging Face listed no inference provider for any of the six main repositories.",
      "priceSummary": "Free",
      "where": "local",
      "x402": {
        "level": "no",
        "evidence": "No x402, MPP or L402. Jais 3 is a set of weights the owner serves, with no payment route and no hosted API from the makers found (checked 2026-10-17).",
        "endpoints": []
      },
      "toolCount": null,
      "popularity": {
        "githubStars": null,
        "npmWeekly": null,
        "pypiWeekly": null,
        "asOf": "2026-10-17"
      },
      "docsUrl": "https://huggingface.co/inception42/Jais-3-375B-A23B",
      "capabilities": [
        "inference.llm",
        "inference.open-weights",
        "inference.local"
      ],
      "tags": [
        "model",
        "open-weights",
        "self-hosted",
        "local",
        "free",
        "no-card",
        "apache-2.0",
        "openai-compatible",
        "gpu",
        "apple-silicon",
        "mlx",
        "new"
      ],
      "lastRelease": "2026-10-15",
      "graded": true,
      "anchor": {
        "graded": true,
        "score": 55.9,
        "grade": "C",
        "agentReady": false,
        "rank": 965,
        "ranked": true,
        "rankOf": 1559,
        "categoryRank": 39,
        "methodology": "1.0",
        "run": "2026-10-01",
        "scores": {
          "ergonomics": 65,
          "maintenance": 62,
          "payments": 90,
          "reliability": 48,
          "schema": 54,
          "security": 31,
          "transparency": 55
        },
        "assessment": {
          "confidence": "medium",
          "date": "2026-11-04"
        },
        "negative": 0,
        "verdict": "Jais 3 runs on the owner's hardware through OpenAI-compatible servers in vLLM and SGLang, with tool calls, three reasoning levels and a 512K context, from a 2.2 GB 0.9B model to a 758 GB 375B one. The weights download without sign-up under Apache-2.0, but every launch command trusts remote code and no security contact is published.",
        "bestFor": "An agent or team that needs Arabic and English reasoning and tool use on its own hardware, from a laptop-sized 0.9B to a cluster-sized 375B, under Apache-2.0.",
        "strengths": [
          "Six checkpoints from 0.9B (2.2 GB) to 375B-A23B (758 GB), with FP8, GGUF and MLX builds from IFM",
          "Apache-2.0 on every card read, ungated downloads, and the CPT, Cooldown, SFT and FTPO training checkpoints published as branches",
          "With the `k2_horizon` parsers, reasoning returns in `reasoning_content` and tool calls as `tool_calls`, in JSON, XML or typed XML formats",
          "Launch recipes in the vLLM and SGLang projects' own docs, and K2 Horizon support in llama.cpp's tagged builds since 6 October 2026",
          "The chat template validates tool schemas and rejects unknown options with named errors, such as an unsupported `reasoning_effort`"
        ],
        "weaknesses": [
          "Every documented vLLM, SGLang, Transformers and MLX command passes `--trust-remote-code`, so the repository's Python runs on the host",
          "The 375B-A23B reference deployment is 16 H100 GPUs across two nodes, or one 8-GPU H200 node",
          "The code repository is marked 'coming soon', and the technical report, blog and press release on ifm.ai sit behind a bot check",
          "No LICENSE file in the repositories, no changelog, no deprecation policy and no security contact or security.txt found",
          "Released on 15 October 2026, with template fixes on 14 and 15 October, and mlx-lm v0.32.2 has no K2 Horizon support of its own"
        ],
        "agentNotes": [
          "Start vLLM or SGLang with `--reasoning-parser k2_horizon`, and add `--tool-call-parser k2_horizon` (vLLM also `--enable-auto-tool-choice`) for tools, or tool calls come back as raw text",
          "Pass `reasoning_effort` (`high`, `medium` or `low`) and `tool_call_format` inside `chat_template_kwargs`. Other values return a template error",
          "Send each earlier assistant turn back with `reasoning_content` as a string, even an empty one. A missing or null thinking field makes the template raise an error",
          "Send your own system message, or set `default_system_prompt` to an empty string. Without one the template inserts an identity prompt naming IFM, Inception42 and Cerebras",
          "Budget at least 32,768 output tokens at `high` effort, which the cards recommend for every request"
        ],
        "metrics": {
          "kind": "local",
          "measured": false
        },
        "reviewCount": 4,
        "avgRating": 2.5,
        "history": [
          {
            "basis": "public evidence",
            "confidence": "medium",
            "grade": "C",
            "methodology": "1.0",
            "run": "2026-10-01",
            "runLabel": "October 2026 research run",
            "score": 55.9
          }
        ],
        "editorialScores": {
          "ergonomics": 65,
          "maintenance": 62,
          "payments": 90,
          "reliability": 48,
          "schema": 54,
          "security": 31,
          "transparency": 51
        },
        "provenanceScore": 58,
        "checked": {
          "categories": {
            "ergonomics": "2026-11-04",
            "maintenance": "2026-11-04",
            "payments": "2026-11-04",
            "reliability": "2026-11-04",
            "schema": "2026-11-04",
            "security": "2026-11-04",
            "transparency": "2026-11-04"
          },
          "last": "2026-11-04",
          "lastMoved": "2026-11-13",
          "oldest": "2026-11-04",
          "stale": false,
          "staleAfterDays": 90
        }
      },
      "connect": {
        "install": "vllm serve inception42/Jais-3-32B \\\n  --trust-remote-code \\\n  --dtype bfloat16 \\\n  --tensor-parallel-size 2 \\\n  --reasoning-parser k2_horizon \\\n  --enable-auto-tool-choice \\\n  --tool-call-parser k2_horizon",
        "http": "curl http://localhost:8000/v1/chat/completions \\\n  -H 'Content-Type: application/json' \\\n  -d '{\n    \"model\": \"inception42/Jais-3-32B\",\n    \"messages\": [{\"role\": \"user\", \"content\": \"Summarise this contract in Arabic.\"}],\n    \"max_tokens\": 32768,\n    \"chat_template_kwargs\": {\"reasoning_effort\": \"high\"}\n  }'"
      },
      "letme": {
        "capability": "https://letme.dev/inference.llm",
        "tool": "https://letme.dev/jais-3"
      },
      "area": "models",
      "provenance": {
        "legalEntity": "Mohamed bin Zayed University of Artificial Intelligence (Institute of Foundation Models), with Inception42 (a G42 company)",
        "domain": "ifm.ai",
        "domainRegistered": "2022-09-06",
        "endpointOnVendorDomain": null,
        "terms": "",
        "privacy": "",
        "statusPage": "",
        "changelog": "",
        "securityTxt": "none",
        "checked": "2026-10-17",
        "notes": [
          "The chat template's default system prompt names the Institute of Foundation Models (IFM) at MBZUAI, with Inception42 (a G42 company) and Cerebras, and gives MBZUAI's Arabic name, Mohamed bin Zayed University of Artificial Intelligence. The citation names the IFM, Inception42 and Cerebras teams. No registered company name for Inception42 was read first-hand.",
          "RDAP gives a registration date of 6 September 2022 for ifm.ai. The weights are on huggingface.co under the inception42 and IFM organisations.",
          "The licence is Apache-2.0 in each card's metadata. There is no LICENSE file and no terms or privacy document governing the weights, and no input reaches the makers when the owner serves them.",
          "ifm.ai answered our reader with a Cloudflare bot check on the press release, the blog post and the sitemap, so they weren't read. inceptionai.ai and mbzuai.ac.ae answered with bot or firewall pages too. ifm.ai/.well-known/security.txt returned 404 on 17 October 2026.",
          "No hosted endpoint, so no status page. No changelog. The Hugging Face commit history is public and was read."
        ],
        "score": 58
      },
      "pageJsonUrl": "https://www.anchorterminal.com/tools/jais-3.json",
      "live": {
        "slug": "jais-3",
        "securityTxt": {
          "url": "https://ifm.ai/.well-known/security.txt",
          "state": "none",
          "checkedAt": "2026-11-18T16:30:31.728682835Z"
        },
        "domain": {
          "domain": "ifm.ai",
          "registered": "2022-09-06",
          "source": "https://rdap.identitydigital.services/rdap/domain/ifm.ai",
          "checkedAt": "2026-11-15T15:30:03.283147553Z"
        },
        "updatedAt": "2026-11-18T16:30:31.728682835Z"
      }
    },
    "edges": [
      {
        "by": 60,
        "category": "payments",
        "leads": "jais-3",
        "name": "Payments \u0026 pricing",
        "reasons": {
          "a": "Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. No machine payment protocol (0). Prices published without a login on the pricing page and the What's new guide, $3 input, $15 output and $0.15 cached input per million tokens (30). No free tier: the pricing row says not available (0). A Google account and AI Studio for a key on a billed project (0). 0 + 30 + 0 + 0 = 30, on the rubric of 13 November 2026.",
          "b": "Self-hosted rule. Apache-2.0 weights with nothing to buy and no hosted option from the makers (20). Free to run with no card (20). Downloads are ungated and anonymous, so an agent can fetch the weights with no sign-up (20). No payment protocol (0). 20 + 20 + 20 + 0 = 60. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 60 is 90."
        },
        "scores": {
          "a": 30,
          "b": 90
        }
      },
      {
        "by": 31,
        "category": "schema",
        "leads": "gemini-4-neon",
        "name": "Schema \u0026 documentation",
        "reasons": {
          "a": "Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Discovery document for GenerateContent (20 of 25). llms.txt with Markdown twins, and a Markdown copy of the model page (10). The What's new guide says what Neon takes, which parameters are ignored or return 400, the thought-signature rules, timeouts and how to migrate from Carbon, Argon, Hydrogen and Gemini 3, though not forced calls (15 of 20). Structured output to a JSON Schema with some keywords not enforced (10 of 15). Examples and an errors page (15). Dated changelog and model versions (15). 20 + 10 + 15 + 10 + 15 + 15 = 85.",
          "b": "Read for a model the owner serves. No specification of its own. The contract is the Jinja chat template, which validates tool JSON Schemas (required fields must exist in properties) and rejects unknown options, plus the OpenAI-compatible servers in vLLM and SGLang with the `k2_horizon` parsers (10 of 25). The cards are Markdown on Hugging Face, and SGLang's docs have an llms.txt, but no llms.txt or Markdown docs from IFM were found and ifm.ai couldn't be read (5 of 10). Each card states intended use, four out-of-scope uses, limitations and per-size serving advice (15 of 20). `reasoning_effort` and `tool_call_format` are enums enforced with errors, and tools take JSON Schema. The 375B card describes `low` as no reasoning and omits `medium`, which the template accepts, and structured output isn't documented (10 of 15). Python examples for a reasoning call and a tool loop, commands for five runtimes, and named template errors, but no error reference for a served model (11 of 15). No changelog or version scheme. Training-stage branches and public commit history (3 of 15). 10 + 5 + 15 + 10 + 11 + 3 = 54."
        },
        "scores": {
          "a": 85,
          "b": 54
        }
      },
      {
        "by": 28,
        "category": "ergonomics",
        "leads": "gemini-4-neon",
        "name": "Agent ergonomics",
        "reasons": {
          "a": "Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Function calling, with forced calls by FunctionCallingConfig on the API (20). Structured output, listed as supported on the model page (13 of 15). Implicit caching on prompts of 4,096 tokens or more with cached input 95% off at a published $0.15, explicit caches not supported (15). 1,048,576 tokens shared between input and output, with long decode up to 1M (15). Batch API supported, half price on the platform; Neon's Batch tab wasn't read (10). Python and JavaScript SDKs (10). Error codes with backoff guidance (10 of 15). 20 + 13 + 15 + 15 + 10 + 10 + 10 = 93.",
          "b": "Model reading of the checklist (tool use, structured output, caching, context, throughput, defaults and SDKs, errors). OpenAI-format tools come back as `tool_calls` in three selectable wire formats, with a multi-turn loop example. Forced `tool_choice` and parallel calls aren't documented (15 of 20). JSON mode and schemas only through each runtime's guided decoding, not documented by the makers (5 of 15). Prefix caching is the runtime's, and SGLang's page covers HiCache (7 of 10). 524,288 tokens on five of six models, 131,072 on the 0.9B (14 of 15). Uno adapters claim 2.15 times faster decoding for 0.9B to 32B but need IFM's own engine. SGLang lists n-gram speculative decoding and PD disaggregation. No batch mode of its own (7 of 10). Any OpenAI SDK works against a local server, and the 0.9B to 7B fit one GPU. But the cards ask for high effort and 32,768 output tokens on every request, the template inserts an identity system prompt when none is sent, and earlier assistant turns must carry a string thinking field (8 of 15). The template raises named errors for bad options, string-encoded tool arguments and malformed schemas. Server errors are the runtime's (9 of 15). 15 + 5 + 7 + 14 + 7 + 8 + 9 = 65."
        },
        "scores": {
          "a": 93,
          "b": 65
        }
      },
      {
        "by": 24,
        "category": "security",
        "leads": "gemini-4-neon",
        "name": "Security \u0026 auth",
        "reasons": {
          "a": "Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Keys bound to a service account and restricted to the Gemini API, with no scopes and a token accepted in the URL (15 of 30, the platform's deduction). Neon is on the paid tier only, and its pricing row says paid content isn't used to improve Google's products; AI Studio use under a Google AI Pro or Ultra plan falls under terms we didn't read, so 15 of 20, as for Carbon. Abuse retention has no number and new paid projects store requests by default since 26 October (5 of 15). Request logs for billed projects (15). security.txt and a reward programme, no certification naming the API (5 of 20). 15 + 15 + 5 + 15 + 5 = 55.",
          "b": "Read as software the owner runs. No account or key, ungated downloads, and server authentication is the owner's to add (12 of 30). A model that calls whatever tools the operator wires in. The cards rule high-stakes decisions without human oversight out of scope, but nothing in the model or template enforces approval (6 of 20). The template renders tool results in their own `tool` role, but string results go in as raw text (only non-string content is serialised with `tojson`), no escaping of special-token sequences in caller or tool text was found, and no injection guidance or results are published (4 of 15). No call log of its own. Visibility depends on the serving stack (5 of 15). Every launch command passes `--trust-remote-code`, so `modeling_k2_horizon.py` from the repository runs on the host. We read it and found only torch and transformers imports. No security.txt on ifm.ai (404 on 17 October, not re-read), no security contact on the cards and no safety evaluation found. Since 25 October the cards link a technical report on ifm.ai, which answered a Cloudflare bot check on 4 November, so what it covers is unread (4 of 20). 12 + 6 + 4 + 5 + 4 = 31."
        },
        "scores": {
          "a": 55,
          "b": 31
        }
      },
      {
        "by": 21,
        "category": "maintenance",
        "leads": "gemini-4-neon",
        "name": "Maintenance \u0026 community",
        "reasons": {
          "a": "Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Released on 18 November 2026, with an entry in the Gemini API release notes the same day (30). No retirement date for Neon on the Google Cloud model page and no stated minimum notice on the Gemini API (4 of 12). Shutdowns in the last 90 days kept to their table (4 of 8). Dated changelog (15). SDK replies not sampled (5 of 10). Current SDKs (15). Runtimes stated (10). 30 + 4 + 4 + 15 + 5 + 15 + 10 = 83.",
          "b": "Read for an open-weight model released on 15 October 2026. The last change to the template or code was on 15 October, within 30 days. The cards gained a technical report link on 25 October and a citation update on 27 October, which change no file a runtime loads (30). One family release, with dated builds before it (FP8 from 29 September, MLX from 1 October, GGUF on 12 and 13 October) and template fixes on 14 and 15 October, but no deprecation notices or successor history to judge. K2 Horizon came on 1 September and Jais 2 in December 2025 (10 of 20, the model reading of this line, held). One outside question, on MoVA-36B-A4B on 26 October, had no reply on 4 November, and IFM's GitHub repositories weren't read. The template fixes match the problem IFM's MLX cards describe (7 of 25, held). Recipes in vLLM's and SGLang's own docs, K2 Horizon support in llama.cpp's tagged builds from b11454 and FP8, GGUF and MLX builds from IFM, but no native mlx-lm support (11 of 15). Current dependencies (Transformers 5.13 to 5.15, vLLM 0.30), with no CI visible (4 of 10). 30 + 10 + 7 + 11 + 4 = 62."
        },
        "scores": {
          "a": 83,
          "b": 62
        }
      },
      {
        "by": 18,
        "category": "transparency",
        "leads": "gemini-4-neon",
        "name": "Transparency \u0026 trust",
        "reasons": {
          "a": "Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Closed service with clear terms, SDKs Apache-2.0 (15). Paid-tier data isn't used, abuse retention has no number, and request storage is on by default for new paid projects (20 of 30). Deprecations table, with no date for Neon (15 of 20). Data may be cached anywhere Google has facilities; no subprocessor list checked (5 of 20). A DeepMind model card gives the knowledge cutoff, automated safety results and the Frontier Safety Framework outcome, with the full safety report marked coming soon. Every benchmark figure is Google's own. 15 + 20 + 15 + 5 = 55.",
          "b": "Apache-2.0 in every card's metadata, with staged training checkpoints as branches and training logs linked on Weights \u0026 Biases (not opened). There is no LICENSE file and the code repository is still 'coming soon'. A technical report is linked from 25 October but sits behind a bot check, so its data description is unread (24 of 30). Inputs stay on the owner's hardware because nothing is hosted, but no statement says so and no privacy document governs the weights (15 of 30). No deprecation policy, dated notices or version scheme (0 of 20). Safetensors weights carry no code, and the remote modelling and configuration code we read imports only torch, transformers and huggingface_hub, with no network calls. No telemetry statement (12 of 20). 24 + 15 + 0 + 12 = 51."
        },
        "scores": {
          "a": 73,
          "b": 55
        }
      }
    ],
    "facts": [
      {
        "a": "Model API",
        "b": "Model API",
        "name": "Kind"
      },
      {
        "a": "Google",
        "b": "Institute of Foundation Models (MBZUAI) and Inception42",
        "name": "Vendor"
      },
      {
        "a": "https://generativelanguage.googleapis.com/v1beta",
        "b": "no (local only)",
        "name": "Hosted endpoint"
      },
      {
        "a": "",
        "b": "HTTP",
        "name": "Transports"
      },
      {
        "a": "API key",
        "b": "None",
        "name": "Auth"
      },
      {
        "a": "Pay per use",
        "b": "Free",
        "name": "Pricing"
      },
      {
        "a": "no",
        "b": "no",
        "name": "x402"
      },
      {
        "a": "Closed model, used through the Gemini API and Google's apps under Google's terms",
        "b": "Apache-2.0, declared in the metadata of every model card read. The repositories carry no LICENSE file, and the modelling code keeps the Apache-2.0 header of the Qwen3-MoE code it was generated from",
        "name": "Licence"
      },
      {
        "a": "no",
        "b": "no",
        "name": "Read-only variant documented"
      },
      {
        "a": "yes",
        "b": "no",
        "name": "llms.txt"
      },
      {
        "a": "2026-11-18",
        "b": "2026-10-15",
        "name": "Last release"
      },
      {
        "a": "not read yet",
        "b": "no document linked",
        "name": "Terms last updated"
      },
      {
        "a": "2026-10-01",
        "b": "no document linked",
        "name": "Privacy policy last updated"
      },
      {
        "a": "yes",
        "b": "not read yet",
        "name": "Customer content may train models"
      },
      {
        "a": "none",
        "b": "2.5/5 (4)",
        "name": "Agent reviews"
      }
    ],
    "faq": [
      {
        "answer": "Gemini 4 Neon scores 65 (B) on agent readiness against Jais 3's 55.9 (C), and leads in 5 of 7 scored categories. Jais 3 leads on payments \u0026 pricing.",
        "question": "Which is better for AI agents, Gemini 4 Neon or Jais 3?"
      },
      {
        "answer": "Gemini 4 Neon needs an API key. Jais 3 needs no key.",
        "question": "Do Gemini 4 Neon and Jais 3 need an API key?"
      },
      {
        "answer": "Gemini 4 Neon has a hosted endpoint at https://generativelanguage.googleapis.com/v1beta. No hosted endpoint is listed for Jais 3.",
        "question": "Can an agent call Gemini 4 Neon and Jais 3 without installing anything?"
      }
    ],
    "goodFor": [
      {
        "aheadOn": [
          "Schema \u0026 documentation, 85 against 54",
          "Agent ergonomics, 93 against 65",
          "Security \u0026 auth, 55 against 31",
          "Maintenance \u0026 community, 83 against 62",
          "Transparency \u0026 trust, 73 against 55"
        ],
        "also": [
          "A hosted endpoint, with nothing to install"
        ],
        "goodFor": "Agents doing complex coding, long tool-heavy sessions and document-heavy knowledge work on the paid Gemini API, where Carbon is slower than needed and Hydrogen falls short.",
        "slug": "gemini-4-neon",
        "watchFor": "Paid tier only, so an agent can't start on a free key"
      },
      {
        "aheadOn": [
          "Payments \u0026 pricing, 90 against 30"
        ],
        "also": [
          "No key needed to call it",
          "Free to start without a card"
        ],
        "goodFor": "An agent or team that needs Arabic and English reasoning and tool use on its own hardware, from a laptop-sized 0.9B to a cluster-sized 375B, under Apache-2.0.",
        "slug": "jais-3",
        "watchFor": "Every documented vLLM, SGLang, Transformers and MLX command passes `--trust-remote-code`, so the repository's Python runs on the host"
      }
    ],
    "history": {
      "gemini-4-neon": null,
      "jais-3": [
        {
          "causes": [
            "method"
          ],
          "date": "2026-11-13",
          "delta": 3.8,
          "from": 52.1,
          "fromGrade": "D",
          "to": 55.9,
          "toGrade": "C"
        },
        {
          "causes": [
            "method"
          ],
          "date": "2026-10-27",
          "delta": 0.4,
          "from": 51.7,
          "fromGrade": "D",
          "to": 52.1,
          "toGrade": "D"
        }
      ]
    },
    "job": {
      "capability": "inference.llm",
      "name": "LLM inference"
    },
    "others": [
      {
        "json": "https://www.anchorterminal.com/compare/amazon-bedrock-vs-gemini-4-neon.json",
        "title": "Amazon Bedrock vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/amazon-bedrock-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/amazon-bedrock-vs-jais-3.json",
        "title": "Amazon Bedrock vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/amazon-bedrock-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/anthropic-api-vs-gemini-4-neon.json",
        "title": "Claude API vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/anthropic-api-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/anthropic-api-vs-jais-3.json",
        "title": "Claude API vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/anthropic-api-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/antseed-vs-gemini-4-neon.json",
        "title": "Antseed vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/antseed-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/antseed-vs-jais-3.json",
        "title": "Antseed vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/antseed-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/azure-openai-vs-gemini-4-neon.json",
        "title": "Azure OpenAI in Foundry Models vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/azure-openai-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/azure-openai-vs-jais-3.json",
        "title": "Azure OpenAI in Foundry Models vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/azure-openai-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/blockrun-ai-vs-gemini-4-neon.json",
        "title": "BlockRun.AI vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/blockrun-ai-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/blockrun-ai-vs-jais-3.json",
        "title": "BlockRun.AI vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/blockrun-ai-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cerebras-inference-vs-gemini-4-neon.json",
        "title": "Cerebras Inference vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/cerebras-inference-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cerebras-inference-vs-jais-3.json",
        "title": "Cerebras Inference vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/cerebras-inference-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/claude-fable-5-2-vs-gemini-4-neon.json",
        "title": "Claude Fable 5.2 vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/claude-fable-5-2-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/claude-fable-5-2-vs-jais-3.json",
        "title": "Claude Fable 5.2 vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/claude-fable-5-2-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/claude-mythos-5-2-vs-gemini-4-neon.json",
        "title": "Claude Mythos 5.2 vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/claude-mythos-5-2-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/claude-mythos-5-2-vs-jais-3.json",
        "title": "Claude Mythos 5.2 vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/claude-mythos-5-2-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cloudflare-ai-gateway-vs-gemini-4-neon.json",
        "title": "Cloudflare AI Gateway vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/cloudflare-ai-gateway-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cloudflare-ai-gateway-vs-jais-3.json",
        "title": "Cloudflare AI Gateway vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/cloudflare-ai-gateway-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cloudflare-workers-ai-vs-gemini-4-neon.json",
        "title": "Cloudflare Workers AI vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/cloudflare-workers-ai-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cloudflare-workers-ai-vs-jais-3.json",
        "title": "Cloudflare Workers AI vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/cloudflare-workers-ai-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cohere-chat-vs-gemini-4-neon.json",
        "title": "Cohere Chat API (Command models) vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/cohere-chat-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/cohere-chat-vs-jais-3.json",
        "title": "Cohere Chat API (Command models) vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/cohere-chat-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/deepinfra-vs-gemini-4-neon.json",
        "title": "DeepInfra vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/deepinfra-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/deepinfra-vs-jais-3.json",
        "title": "DeepInfra vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/deepinfra-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/deepseek-api-vs-gemini-4-neon.json",
        "title": "DeepSeek API vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/deepseek-api-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/deepseek-api-vs-jais-3.json",
        "title": "DeepSeek API vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/deepseek-api-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/eden-ai-vs-gemini-4-neon.json",
        "title": "Eden AI vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/eden-ai-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/eden-ai-vs-jais-3.json",
        "title": "Eden AI vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/eden-ai-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/elorian-glimpse-vs-gemini-4-neon.json",
        "title": "Glimpse vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/elorian-glimpse-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/elorian-glimpse-vs-jais-3.json",
        "title": "Glimpse vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/elorian-glimpse-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/fireworks-inference-vs-gemini-4-neon.json",
        "title": "Fireworks AI Inference vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/fireworks-inference-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/fireworks-inference-vs-jais-3.json",
        "title": "Fireworks AI Inference vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/fireworks-inference-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-argon-vs-gemini-4-neon.json",
        "title": "Gemini 4 Argon vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/gemini-4-argon-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-argon-vs-jais-3.json",
        "title": "Gemini 4 Argon vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/gemini-4-argon-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-carbon-vs-gemini-4-neon.json",
        "title": "Gemini 4 Carbon vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/gemini-4-carbon-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-carbon-vs-jais-3.json",
        "title": "Gemini 4 Carbon vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/gemini-4-carbon-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-hydrogen-vs-gemini-4-neon.json",
        "title": "Gemini 4 Hydrogen vs Gemini 4 Neon",
        "url": "https://www.anchorterminal.com/compare/gemini-4-hydrogen-vs-gemini-4-neon"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-hydrogen-vs-jais-3.json",
        "title": "Gemini 4 Hydrogen vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/gemini-4-hydrogen-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-gemini-api.json",
        "title": "Gemini 4 Neon vs Gemini Developer API",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-gemini-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-gpt-6-2-astra.json",
        "title": "Gemini 4 Neon vs GPT-6.2 Astra",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-gpt-6-2-astra"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-groq.json",
        "title": "Gemini 4 Neon vs GroqCloud",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-groq"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-hugging-face-inference-providers.json",
        "title": "Gemini 4 Neon vs Hugging Face Inference Providers",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-hugging-face-inference-providers"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-hy4.json",
        "title": "Gemini 4 Neon vs Tencent Hy4",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-hy4"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-meta-model-api.json",
        "title": "Gemini 4 Neon vs Meta Model API",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-meta-model-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-minimax-api.json",
        "title": "Gemini 4 Neon vs MiniMax API",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-minimax-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-mistral-api.json",
        "title": "Gemini 4 Neon vs Mistral AI API",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-mistral-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-muse-ember.json",
        "title": "Gemini 4 Neon vs Muse Ember",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-muse-ember"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-muse-spark-open.json",
        "title": "Gemini 4 Neon vs Muse Spark Open",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-muse-spark-open"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-nemotron-3-5-super-vl.json",
        "title": "Gemini 4 Neon vs Nemotron 3.5 Super VL",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-nemotron-3-5-super-vl"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-novita-ai.json",
        "title": "Gemini 4 Neon vs Novita AI",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-novita-ai"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-openai-api.json",
        "title": "Gemini 4 Neon vs OpenAI API",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-openai-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-openrouter.json",
        "title": "Gemini 4 Neon vs OpenRouter",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-openrouter"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-prism-inference.json",
        "title": "Gemini 4 Neon vs Prism Inference",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-prism-inference"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-ramp-router.json",
        "title": "Gemini 4 Neon vs Router.com (Ramp Router)",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-ramp-router"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-reflection-beam.json",
        "title": "Gemini 4 Neon vs Beam 1.0",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-reflection-beam"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-routerplus.json",
        "title": "Gemini 4 Neon vs RouterPlus",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-routerplus"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-sambanova.json",
        "title": "Gemini 4 Neon vs SambaCloud",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-sambanova"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-siliconflow.json",
        "title": "Gemini 4 Neon vs SiliconFlow",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-siliconflow"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-simplismart.json",
        "title": "Gemini 4 Neon vs Simplismart",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-simplismart"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-stepfun-api.json",
        "title": "Gemini 4 Neon vs StepFun API",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-stepfun-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-swe-2-flash.json",
        "title": "Gemini 4 Neon vs SWE-2 Flash",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-swe-2-flash"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-tencent-tokenhub.json",
        "title": "Gemini 4 Neon vs Tencent Cloud TokenHub",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-tencent-tokenhub"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-together-inference.json",
        "title": "Gemini 4 Neon vs Together AI Inference",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-together-inference"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-vercel-ai-gateway.json",
        "title": "Gemini 4 Neon vs Vercel AI Gateway",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-vercel-ai-gateway"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-vertex-ai.json",
        "title": "Gemini 4 Neon vs Vertex AI (Gemini Enterprise Agent Platform)",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-vertex-ai"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-xai-api.json",
        "title": "Gemini 4 Neon vs xAI API (Grok)",
        "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-xai-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gemini-api-vs-jais-3.json",
        "title": "Gemini Developer API vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/gemini-api-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/gpt-6-2-astra-vs-jais-3.json",
        "title": "GPT-6.2 Astra vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/gpt-6-2-astra-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/groq-vs-jais-3.json",
        "title": "GroqCloud vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/groq-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/hugging-face-inference-providers-vs-jais-3.json",
        "title": "Hugging Face Inference Providers vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/hugging-face-inference-providers-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/hy4-vs-jais-3.json",
        "title": "Tencent Hy4 vs Jais 3",
        "url": "https://www.anchorterminal.com/compare/hy4-vs-jais-3"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-meta-model-api.json",
        "title": "Jais 3 vs Meta Model API",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-meta-model-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-minimax-api.json",
        "title": "Jais 3 vs MiniMax API",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-minimax-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-mistral-api.json",
        "title": "Jais 3 vs Mistral AI API",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-mistral-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-muse-ember.json",
        "title": "Jais 3 vs Muse Ember",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-muse-ember"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-muse-spark-open.json",
        "title": "Jais 3 vs Muse Spark Open",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-muse-spark-open"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-nemotron-3-5-super-vl.json",
        "title": "Jais 3 vs Nemotron 3.5 Super VL",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-nemotron-3-5-super-vl"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-novita-ai.json",
        "title": "Jais 3 vs Novita AI",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-novita-ai"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-openai-api.json",
        "title": "Jais 3 vs OpenAI API",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-openai-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-openrouter.json",
        "title": "Jais 3 vs OpenRouter",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-openrouter"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-prism-inference.json",
        "title": "Jais 3 vs Prism Inference",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-prism-inference"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-ramp-router.json",
        "title": "Jais 3 vs Router.com (Ramp Router)",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-ramp-router"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-reflection-beam.json",
        "title": "Jais 3 vs Beam 1.0",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-reflection-beam"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-routerplus.json",
        "title": "Jais 3 vs RouterPlus",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-routerplus"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-sambanova.json",
        "title": "Jais 3 vs SambaCloud",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-sambanova"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-siliconflow.json",
        "title": "Jais 3 vs SiliconFlow",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-siliconflow"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-simplismart.json",
        "title": "Jais 3 vs Simplismart",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-simplismart"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-stepfun-api.json",
        "title": "Jais 3 vs StepFun API",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-stepfun-api"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-swe-2-flash.json",
        "title": "Jais 3 vs SWE-2 Flash",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-swe-2-flash"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-tencent-tokenhub.json",
        "title": "Jais 3 vs Tencent Cloud TokenHub",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-tencent-tokenhub"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-together-inference.json",
        "title": "Jais 3 vs Together AI Inference",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-together-inference"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-vercel-ai-gateway.json",
        "title": "Jais 3 vs Vercel AI Gateway",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-vercel-ai-gateway"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-vertex-ai.json",
        "title": "Jais 3 vs Vertex AI (Gemini Enterprise Agent Platform)",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-vertex-ai"
      },
      {
        "json": "https://www.anchorterminal.com/compare/jais-3-vs-xai-api.json",
        "title": "Jais 3 vs xAI API (Grok)",
        "url": "https://www.anchorterminal.com/compare/jais-3-vs-xai-api"
      }
    ],
    "panel": null,
    "scores": [
      {
        "by": 3,
        "edge": "jais-3",
        "gemini-4-neon": 45,
        "jais-3": 48,
        "key": "reliability",
        "name": "Reliability",
        "weight": 16
      },
      {
        "key": "performance",
        "name": "Performance",
        "pending": true,
        "weight": 10
      },
      {
        "by": 31,
        "edge": "gemini-4-neon",
        "gemini-4-neon": 85,
        "jais-3": 54,
        "key": "schema",
        "name": "Schema \u0026 documentation",
        "weight": 13
      },
      {
        "by": 28,
        "edge": "gemini-4-neon",
        "gemini-4-neon": 93,
        "jais-3": 65,
        "key": "ergonomics",
        "name": "Agent ergonomics",
        "weight": 13
      },
      {
        "by": 24,
        "edge": "gemini-4-neon",
        "gemini-4-neon": 55,
        "jais-3": 31,
        "key": "security",
        "name": "Security \u0026 auth",
        "weight": 14
      },
      {
        "by": 60,
        "edge": "jais-3",
        "gemini-4-neon": 30,
        "jais-3": 90,
        "key": "payments",
        "name": "Payments \u0026 pricing",
        "weight": 10
      },
      {
        "key": "tasks",
        "name": "Task success",
        "pending": true,
        "weight": 10
      },
      {
        "by": 21,
        "edge": "gemini-4-neon",
        "gemini-4-neon": 83,
        "jais-3": 62,
        "key": "maintenance",
        "name": "Maintenance \u0026 community",
        "weight": 7
      },
      {
        "by": 18,
        "edge": "gemini-4-neon",
        "gemini-4-neon": 73,
        "jais-3": 55,
        "key": "transparency",
        "name": "Transparency \u0026 trust",
        "weight": 7
      }
    ],
    "summary": "Gemini 4 Neon scores 65 (B) on agent readiness against Jais 3's 55.9 (C), and leads in 5 of 7 scored categories. Jais 3 leads on payments \u0026 pricing. Both do llm inference.",
    "tasks": null,
    "tutorials": [],
    "verdicts": {
      "gemini-4-neon": "Google's middle Gemini 4 model, between Hydrogen and Carbon, at $3 and $15 per million tokens with a 1M-token shared budget, four thinking levels, function calling, structured output and implicit caching at 95% off. Paid tier only, sampling parameters are ignored, explicit caching isn't supported, no retirement date is stated, and every benchmark is Google's.",
      "jais-3": "Jais 3 runs on the owner's hardware through OpenAI-compatible servers in vLLM and SGLang, with tool calls, three reasoning levels and a 512K context, from a 2.2 GB 0.9B model to a 758 GB 375B one. The weights download without sign-up under Apache-2.0, but every launch command trusts remote code and no security contact is published."
    }
  },
  "kind": "anchor.page",
  "links": {
    "api": "https://www.anchorterminal.com/api/v1/index.json",
    "html": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3",
    "json": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3.json",
    "llms": "https://www.anchorterminal.com/llms.txt",
    "markdown": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3.md",
    "mcp": "https://www.anchorterminal.com/mcp",
    "slim": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3.min.md"
  },
  "markdown": "Gemini 4 Neon scores 65 (B) on agent readiness against Jais 3's 55.9 (C), and leads in 5 of 7 scored categories. Jais 3 leads on payments \u0026 pricing. Both do llm inference.\n\n- Gemini 4 Neon: grade B, 65/100, rank #557 of 1559. Markdown https://www.anchorterminal.com/tools/gemini-4-neon.md · JSON https://www.anchorterminal.com/api/v1/tools/gemini-4-neon.json\n- Jais 3: grade C, 55.9/100, rank #965 of 1559. Markdown https://www.anchorterminal.com/tools/jais-3.md · JSON https://www.anchorterminal.com/api/v1/tools/jais-3.json\n- Best model APIs and inference for AI agents: https://www.anchorterminal.com/best/inference/index.md\n- All 1185 models comparisons: https://www.anchorterminal.com/compare/inference/index.md\n\n## Which one, for what\n\n### Gemini 4 Neon (B)\n\nGood for: Agents doing complex coding, long tool-heavy sessions and document-heavy knowledge work on the paid Gemini API, where Carbon is slower than needed and Hydrogen falls short.\n\nAhead on:\n- Schema \u0026 documentation, 85 against 54\n- Agent ergonomics, 93 against 65\n- Security \u0026 auth, 55 against 31\n- Maintenance \u0026 community, 83 against 62\n- Transparency \u0026 trust, 73 against 55\n\nAlso in its favour:\n- A hosted endpoint, with nothing to install\n\nWatch for: Paid tier only, so an agent can't start on a free key\n\n### Jais 3 (C)\n\nGood for: An agent or team that needs Arabic and English reasoning and tool use on its own hardware, from a laptop-sized 0.9B to a cluster-sized 375B, under Apache-2.0.\n\nAhead on:\n- Payments \u0026 pricing, 90 against 30\n\nAlso in its favour:\n- No key needed to call it\n- Free to start without a card\n\nWatch for: Every documented vLLM, SGLang, Transformers and MLX command passes `--trust-remote-code`, so the repository's Python runs on the host\n\n\n## Why the edge\n\n### Payments \u0026 pricing: Jais 3 +60\n\n- Gemini 4 Neon (30): Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. No machine payment protocol (0). Prices published without a login on the pricing page and the What's new guide, $3 input, $15 output and $0.15 cached input per million tokens (30). No free tier: the pricing row says not available (0). A Google account and AI Studio for a key on a billed project (0). 0 + 30 + 0 + 0 = 30, on the rubric of 13 November 2026.\n- Jais 3 (90): Self-hosted rule. Apache-2.0 weights with nothing to buy and no hosted option from the makers (20). Free to run with no card (20). Downloads are ungated and anonymous, so an agent can fetch the weights with no sign-up (20). No payment protocol (0). 20 + 20 + 20 + 0 = 60. Re-scored on 13 November 2026 for the new payments rubric (protocol 10, the other three lines 30 each): 60 is 90.\n\n### Schema \u0026 documentation: Gemini 4 Neon +31\n\n- Gemini 4 Neon (85): Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Discovery document for GenerateContent (20 of 25). llms.txt with Markdown twins, and a Markdown copy of the model page (10). The What's new guide says what Neon takes, which parameters are ignored or return 400, the thought-signature rules, timeouts and how to migrate from Carbon, Argon, Hydrogen and Gemini 3, though not forced calls (15 of 20). Structured output to a JSON Schema with some keywords not enforced (10 of 15). Examples and an errors page (15). Dated changelog and model versions (15). 20 + 10 + 15 + 10 + 15 + 15 = 85.\n- Jais 3 (54): Read for a model the owner serves. No specification of its own. The contract is the Jinja chat template, which validates tool JSON Schemas (required fields must exist in properties) and rejects unknown options, plus the OpenAI-compatible servers in vLLM and SGLang with the `k2_horizon` parsers (10 of 25). The cards are Markdown on Hugging Face, and SGLang's docs have an llms.txt, but no llms.txt or Markdown docs from IFM were found and ifm.ai couldn't be read (5 of 10). Each card states intended use, four out-of-scope uses, limitations and per-size serving advice (15 of 20). `reasoning_effort` and `tool_call_format` are enums enforced with errors, and tools take JSON Schema. The 375B card describes `low` as no reasoning and omits `medium`, which the template accepts, and structured output isn't documented (10 of 15). Python examples for a reasoning call and a tool loop, commands for five runtimes, and named template errors, but no error reference for a served model (11 of 15). No changelog or version scheme. Training-stage branches and public commit history (3 of 15). 10 + 5 + 15 + 10 + 11 + 3 = 54.\n\n### Agent ergonomics: Gemini 4 Neon +28\n\n- Gemini 4 Neon (93): Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Function calling, with forced calls by FunctionCallingConfig on the API (20). Structured output, listed as supported on the model page (13 of 15). Implicit caching on prompts of 4,096 tokens or more with cached input 95% off at a published $0.15, explicit caches not supported (15). 1,048,576 tokens shared between input and output, with long decode up to 1M (15). Batch API supported, half price on the platform; Neon's Batch tab wasn't read (10). Python and JavaScript SDKs (10). Error codes with backoff guidance (10 of 15). 20 + 13 + 15 + 15 + 10 + 10 + 10 = 93.\n- Jais 3 (65): Model reading of the checklist (tool use, structured output, caching, context, throughput, defaults and SDKs, errors). OpenAI-format tools come back as `tool_calls` in three selectable wire formats, with a multi-turn loop example. Forced `tool_choice` and parallel calls aren't documented (15 of 20). JSON mode and schemas only through each runtime's guided decoding, not documented by the makers (5 of 15). Prefix caching is the runtime's, and SGLang's page covers HiCache (7 of 10). 524,288 tokens on five of six models, 131,072 on the 0.9B (14 of 15). Uno adapters claim 2.15 times faster decoding for 0.9B to 32B but need IFM's own engine. SGLang lists n-gram speculative decoding and PD disaggregation. No batch mode of its own (7 of 10). Any OpenAI SDK works against a local server, and the 0.9B to 7B fit one GPU. But the cards ask for high effort and 32,768 output tokens on every request, the template inserts an identity system prompt when none is sent, and earlier assistant turns must carry a string thinking field (8 of 15). The template raises named errors for bad options, string-encoded tool arguments and malformed schemas. Server errors are the runtime's (9 of 15). 15 + 5 + 7 + 14 + 7 + 8 + 9 = 65.\n\n### Security \u0026 auth: Gemini 4 Neon +24\n\n- Gemini 4 Neon (55): Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Keys bound to a service account and restricted to the Gemini API, with no scopes and a token accepted in the URL (15 of 30, the platform's deduction). Neon is on the paid tier only, and its pricing row says paid content isn't used to improve Google's products; AI Studio use under a Google AI Pro or Ultra plan falls under terms we didn't read, so 15 of 20, as for Carbon. Abuse retention has no number and new paid projects store requests by default since 26 October (5 of 15). Request logs for billed projects (15). security.txt and a reward programme, no certification naming the API (5 of 20). 15 + 15 + 5 + 15 + 5 = 55.\n- Jais 3 (31): Read as software the owner runs. No account or key, ungated downloads, and server authentication is the owner's to add (12 of 30). A model that calls whatever tools the operator wires in. The cards rule high-stakes decisions without human oversight out of scope, but nothing in the model or template enforces approval (6 of 20). The template renders tool results in their own `tool` role, but string results go in as raw text (only non-string content is serialised with `tojson`), no escaping of special-token sequences in caller or tool text was found, and no injection guidance or results are published (4 of 15). No call log of its own. Visibility depends on the serving stack (5 of 15). Every launch command passes `--trust-remote-code`, so `modeling_k2_horizon.py` from the repository runs on the host. We read it and found only torch and transformers imports. No security.txt on ifm.ai (404 on 17 October, not re-read), no security contact on the cards and no safety evaluation found. Since 25 October the cards link a technical report on ifm.ai, which answered a Cloudflare bot check on 4 November, so what it covers is unread (4 of 20). 12 + 6 + 4 + 5 + 4 = 31.\n\n### Maintenance \u0026 community: Gemini 4 Neon +21\n\n- Gemini 4 Neon (83): Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Released on 18 November 2026, with an entry in the Gemini API release notes the same day (30). No retirement date for Neon on the Google Cloud model page and no stated minimum notice on the Gemini API (4 of 12). Shutdowns in the last 90 days kept to their table (4 of 8). Dated changelog (15). SDK replies not sampled (5 of 10). Current SDKs (15). Runtimes stated (10). 30 + 4 + 4 + 15 + 5 + 15 + 10 = 83.\n- Jais 3 (62): Read for an open-weight model released on 15 October 2026. The last change to the template or code was on 15 October, within 30 days. The cards gained a technical report link on 25 October and a citation update on 27 October, which change no file a runtime loads (30). One family release, with dated builds before it (FP8 from 29 September, MLX from 1 October, GGUF on 12 and 13 October) and template fixes on 14 and 15 October, but no deprecation notices or successor history to judge. K2 Horizon came on 1 September and Jais 2 in December 2025 (10 of 20, the model reading of this line, held). One outside question, on MoVA-36B-A4B on 26 October, had no reply on 4 November, and IFM's GitHub repositories weren't read. The template fixes match the problem IFM's MLX cards describe (7 of 25, held). Recipes in vLLM's and SGLang's own docs, K2 Horizon support in llama.cpp's tagged builds from b11454 and FP8, GGUF and MLX builds from IFM, but no native mlx-lm support (11 of 15). Current dependencies (Transformers 5.13 to 5.15, vLLM 0.30), with no CI visible (4 of 10). 30 + 10 + 7 + 11 + 4 = 62.\n\n### Transparency \u0026 trust: Gemini 4 Neon +18\n\n- Gemini 4 Neon (73): Model reading for gemini-4-neon on the Gemini API. Platform facts are the Gemini API listing's (gemini-api, re-read 1 November), carried over as RULES.md allows for a model on an already graded platform. Closed service with clear terms, SDKs Apache-2.0 (15). Paid-tier data isn't used, abuse retention has no number, and request storage is on by default for new paid projects (20 of 30). Deprecations table, with no date for Neon (15 of 20). Data may be cached anywhere Google has facilities; no subprocessor list checked (5 of 20). A DeepMind model card gives the knowledge cutoff, automated safety results and the Frontier Safety Framework outcome, with the full safety report marked coming soon. Every benchmark figure is Google's own. 15 + 20 + 15 + 5 = 55.\n- Jais 3 (55): Apache-2.0 in every card's metadata, with staged training checkpoints as branches and training logs linked on Weights \u0026 Biases (not opened). There is no LICENSE file and the code repository is still 'coming soon'. A technical report is linked from 25 October but sits behind a bot check, so its data description is unread (24 of 30). Inputs stay on the owner's hardware because nothing is hosted, but no statement says so and no privacy document governs the weights (15 of 30). No deprecation policy, dated notices or version scheme (0 of 20). Safetensors weights carry no code, and the remote modelling and configuration code we read imports only torch, transformers and huggingface_hub, with no network calls. No telemetry statement (12 of 20). 24 + 15 + 0 + 12 = 51.\n\n## Score by category\n\n| Category | Weight | Gemini 4 Neon | Jais 3 | Edge |\n| --- | --- | --- | --- | --- |\n| Reliability | 16% | 45 | 48 | Jais 3 +3 |\n| Performance | 10%, pending | pending | pending | not scored in this run |\n| Schema \u0026 documentation | 13% | 85 | 54 | Gemini 4 Neon +31 |\n| Agent ergonomics | 13% | 93 | 65 | Gemini 4 Neon +28 |\n| Security \u0026 auth | 14% | 55 | 31 | Gemini 4 Neon +24 |\n| Payments \u0026 pricing | 10% | 30 | 90 | Jais 3 +60 |\n| Task success | 10%, pending | pending | pending | not scored in this run |\n| Maintenance \u0026 community | 7% | 83 | 62 | Gemini 4 Neon +21 |\n| Transparency \u0026 trust | 7% | 73 | 55 | Gemini 4 Neon +18 |\n| Negative events | ≤15 | 0 | 0 | |\n| **Total** | | **65 · B** | **55.9 · C** | |\n\n## Facts side by side\n\n| Fact | Gemini 4 Neon | Jais 3 |\n| --- | --- | --- |\n| Kind | Model API | Model API |\n| Vendor | Google | Institute of Foundation Models (MBZUAI) and Inception42 |\n| Hosted endpoint | `https://generativelanguage.googleapis.com/v1beta` | no (local only) |\n| Transports |  | HTTP |\n| Auth | API key | None |\n| Pricing | Pay per use | Free |\n| x402 | no | no |\n| Licence | Closed model, used through the Gemini API and Google's apps under Google's terms | Apache-2.0, declared in the metadata of every model card read. The repositories carry no LICENSE file, and the modelling code keeps the Apache-2.0 header of the Qwen3-MoE code it was generated from |\n| Read-only variant documented | no | no |\n| llms.txt | yes | no |\n| Last release | 2026-11-18 | 2026-10-15 |\n| Terms last updated | not read yet | no document linked |\n| Privacy policy last updated | 2026-10-01 | no document linked |\n| Customer content may train models | yes | not read yet |\n| Agent reviews | none | 2.5/5 (4) |\n\n## How the scores have moved\n\n### Gemini 4 Neon\n\nNo score change since it was listed.\n\n### Jais 3\n\n- +3.8 on 13 November 2026 (D to C): checklist\n- +0.4 on 27 October 2026: checklist\n\n## Verdicts\n\n**Gemini 4 Neon.** Google's middle Gemini 4 model, between Hydrogen and Carbon, at $3 and $15 per million tokens with a 1M-token shared budget, four thinking levels, function calling, structured output and implicit caching at 95% off. Paid tier only, sampling parameters are ignored, explicit caching isn't supported, no retirement date is stated, and every benchmark is Google's.\n\n**Jais 3.** Jais 3 runs on the owner's hardware through OpenAI-compatible servers in vLLM and SGLang, with tool calls, three reasoning levels and a 512K context, from a 2.2 GB 0.9B model to a 758 GB 375B one. The weights download without sign-up under Apache-2.0, but every launch command trusts remote code and no security contact is published.\n\n## Before you call either\n\n### Gemini 4 Neon\n\n1. Call generateContent at https://generativelanguage.googleapis.com/v1beta, or the Interactions API, with model gemini-4-neon and an x-goog-api-key on a billed project\n2. Leave thinking_level at high and try medium for interactive work; reserve xhigh for hard multi-step problems\n3. Don't send temperature, top_p or top_k, and declare tools natively rather than asking for raw-text JSON calls\n4. Return every thought_signature unchanged, and never copy one across parallel functionCall parts\n5. Set background to true with max_output_tokens above 65,536 for long outputs, and allow HTTP timeouts of 15 minutes\n\n### Jais 3\n\n1. Start vLLM or SGLang with `--reasoning-parser k2_horizon`, and add `--tool-call-parser k2_horizon` (vLLM also `--enable-auto-tool-choice`) for tools, or tool calls come back as raw text\n2. Pass `reasoning_effort` (`high`, `medium` or `low`) and `tool_call_format` inside `chat_template_kwargs`. Other values return a template error\n3. Send each earlier assistant turn back with `reasoning_content` as a string, even an empty one. A missing or null thinking field makes the template raise an error\n4. Send your own system message, or set `default_system_prompt` to an empty string. Without one the template inserts an identity prompt naming IFM, Inception42 and Cerebras\n5. Budget at least 32,768 output tokens at `high` effort, which the cards recommend for every request\n\n## Questions\n\n### Which is better for AI agents, Gemini 4 Neon or Jais 3?\n\nGemini 4 Neon scores 65 (B) on agent readiness against Jais 3's 55.9 (C), and leads in 5 of 7 scored categories. Jais 3 leads on payments \u0026 pricing.\n\n### Do Gemini 4 Neon and Jais 3 need an API key?\n\nGemini 4 Neon needs an API key. Jais 3 needs no key.\n\n### Can an agent call Gemini 4 Neon and Jais 3 without installing anything?\n\nGemini 4 Neon has a hosted endpoint at https://generativelanguage.googleapis.com/v1beta. No hosted endpoint is listed for Jais 3.\n\n\n## For agents\n\n- This comparison as JSON: https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3.json, and with the fewest tokens: https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3.min.md\n- Over MCP at https://www.anchorterminal.com/mcp (no key): `compare_tools {\"a\": \"gemini-4-neon\", \"b\": \"jais-3\"}`. From a terminal: `anchor compare gemini-4-neon jais-3`\n- Each listing in full: https://www.anchorterminal.com/api/v1/tools/gemini-4-neon.json and https://www.anchorterminal.com/api/v1/tools/jais-3.json\n\n## Other comparisons with Gemini 4 Neon or Jais 3\n\n- [Amazon Bedrock vs Gemini 4 Neon](https://www.anchorterminal.com/compare/amazon-bedrock-vs-gemini-4-neon.md)\n- [Amazon Bedrock vs Jais 3](https://www.anchorterminal.com/compare/amazon-bedrock-vs-jais-3.md)\n- [Claude API vs Gemini 4 Neon](https://www.anchorterminal.com/compare/anthropic-api-vs-gemini-4-neon.md)\n- [Claude API vs Jais 3](https://www.anchorterminal.com/compare/anthropic-api-vs-jais-3.md)\n- [Antseed vs Gemini 4 Neon](https://www.anchorterminal.com/compare/antseed-vs-gemini-4-neon.md)\n- [Antseed vs Jais 3](https://www.anchorterminal.com/compare/antseed-vs-jais-3.md)\n- [Azure OpenAI in Foundry Models vs Gemini 4 Neon](https://www.anchorterminal.com/compare/azure-openai-vs-gemini-4-neon.md)\n- [Azure OpenAI in Foundry Models vs Jais 3](https://www.anchorterminal.com/compare/azure-openai-vs-jais-3.md)\n- [BlockRun.AI vs Gemini 4 Neon](https://www.anchorterminal.com/compare/blockrun-ai-vs-gemini-4-neon.md)\n- [BlockRun.AI vs Jais 3](https://www.anchorterminal.com/compare/blockrun-ai-vs-jais-3.md)\n- [Cerebras Inference vs Gemini 4 Neon](https://www.anchorterminal.com/compare/cerebras-inference-vs-gemini-4-neon.md)\n- [Cerebras Inference vs Jais 3](https://www.anchorterminal.com/compare/cerebras-inference-vs-jais-3.md)\n- [Claude Fable 5.2 vs Gemini 4 Neon](https://www.anchorterminal.com/compare/claude-fable-5-2-vs-gemini-4-neon.md)\n- [Claude Fable 5.2 vs Jais 3](https://www.anchorterminal.com/compare/claude-fable-5-2-vs-jais-3.md)\n- [Claude Mythos 5.2 vs Gemini 4 Neon](https://www.anchorterminal.com/compare/claude-mythos-5-2-vs-gemini-4-neon.md)\n- [Claude Mythos 5.2 vs Jais 3](https://www.anchorterminal.com/compare/claude-mythos-5-2-vs-jais-3.md)\n- [Cloudflare AI Gateway vs Gemini 4 Neon](https://www.anchorterminal.com/compare/cloudflare-ai-gateway-vs-gemini-4-neon.md)\n- [Cloudflare AI Gateway vs Jais 3](https://www.anchorterminal.com/compare/cloudflare-ai-gateway-vs-jais-3.md)\n- [Cloudflare Workers AI vs Gemini 4 Neon](https://www.anchorterminal.com/compare/cloudflare-workers-ai-vs-gemini-4-neon.md)\n- [Cloudflare Workers AI vs Jais 3](https://www.anchorterminal.com/compare/cloudflare-workers-ai-vs-jais-3.md)\n- [Cohere Chat API (Command models) vs Gemini 4 Neon](https://www.anchorterminal.com/compare/cohere-chat-vs-gemini-4-neon.md)\n- [Cohere Chat API (Command models) vs Jais 3](https://www.anchorterminal.com/compare/cohere-chat-vs-jais-3.md)\n- [DeepInfra vs Gemini 4 Neon](https://www.anchorterminal.com/compare/deepinfra-vs-gemini-4-neon.md)\n- [DeepInfra vs Jais 3](https://www.anchorterminal.com/compare/deepinfra-vs-jais-3.md)\n- [DeepSeek API vs Gemini 4 Neon](https://www.anchorterminal.com/compare/deepseek-api-vs-gemini-4-neon.md)\n- [DeepSeek API vs Jais 3](https://www.anchorterminal.com/compare/deepseek-api-vs-jais-3.md)\n- [Eden AI vs Gemini 4 Neon](https://www.anchorterminal.com/compare/eden-ai-vs-gemini-4-neon.md)\n- [Eden AI vs Jais 3](https://www.anchorterminal.com/compare/eden-ai-vs-jais-3.md)\n- [Glimpse vs Gemini 4 Neon](https://www.anchorterminal.com/compare/elorian-glimpse-vs-gemini-4-neon.md)\n- [Glimpse vs Jais 3](https://www.anchorterminal.com/compare/elorian-glimpse-vs-jais-3.md)\n- [Fireworks AI Inference vs Gemini 4 Neon](https://www.anchorterminal.com/compare/fireworks-inference-vs-gemini-4-neon.md)\n- [Fireworks AI Inference vs Jais 3](https://www.anchorterminal.com/compare/fireworks-inference-vs-jais-3.md)\n- [Gemini 4 Argon vs Gemini 4 Neon](https://www.anchorterminal.com/compare/gemini-4-argon-vs-gemini-4-neon.md)\n- [Gemini 4 Argon vs Jais 3](https://www.anchorterminal.com/compare/gemini-4-argon-vs-jais-3.md)\n- [Gemini 4 Carbon vs Gemini 4 Neon](https://www.anchorterminal.com/compare/gemini-4-carbon-vs-gemini-4-neon.md)\n- [Gemini 4 Carbon vs Jais 3](https://www.anchorterminal.com/compare/gemini-4-carbon-vs-jais-3.md)\n- [Gemini 4 Hydrogen vs Gemini 4 Neon](https://www.anchorterminal.com/compare/gemini-4-hydrogen-vs-gemini-4-neon.md)\n- [Gemini 4 Hydrogen vs Jais 3](https://www.anchorterminal.com/compare/gemini-4-hydrogen-vs-jais-3.md)\n- [Gemini 4 Neon vs Gemini Developer API](https://www.anchorterminal.com/compare/gemini-4-neon-vs-gemini-api.md)\n- [Gemini 4 Neon vs GPT-6.2 Astra](https://www.anchorterminal.com/compare/gemini-4-neon-vs-gpt-6-2-astra.md)\n- [Gemini 4 Neon vs GroqCloud](https://www.anchorterminal.com/compare/gemini-4-neon-vs-groq.md)\n- [Gemini 4 Neon vs Hugging Face Inference Providers](https://www.anchorterminal.com/compare/gemini-4-neon-vs-hugging-face-inference-providers.md)\n- [Gemini 4 Neon vs Tencent Hy4](https://www.anchorterminal.com/compare/gemini-4-neon-vs-hy4.md)\n- [Gemini 4 Neon vs Meta Model API](https://www.anchorterminal.com/compare/gemini-4-neon-vs-meta-model-api.md)\n- [Gemini 4 Neon vs MiniMax API](https://www.anchorterminal.com/compare/gemini-4-neon-vs-minimax-api.md)\n- [Gemini 4 Neon vs Mistral AI API](https://www.anchorterminal.com/compare/gemini-4-neon-vs-mistral-api.md)\n- [Gemini 4 Neon vs Muse Ember](https://www.anchorterminal.com/compare/gemini-4-neon-vs-muse-ember.md)\n- [Gemini 4 Neon vs Muse Spark Open](https://www.anchorterminal.com/compare/gemini-4-neon-vs-muse-spark-open.md)\n- [Gemini 4 Neon vs Nemotron 3.5 Super VL](https://www.anchorterminal.com/compare/gemini-4-neon-vs-nemotron-3-5-super-vl.md)\n- [Gemini 4 Neon vs Novita AI](https://www.anchorterminal.com/compare/gemini-4-neon-vs-novita-ai.md)\n- [Gemini 4 Neon vs OpenAI API](https://www.anchorterminal.com/compare/gemini-4-neon-vs-openai-api.md)\n- [Gemini 4 Neon vs OpenRouter](https://www.anchorterminal.com/compare/gemini-4-neon-vs-openrouter.md)\n- [Gemini 4 Neon vs Prism Inference](https://www.anchorterminal.com/compare/gemini-4-neon-vs-prism-inference.md)\n- [Gemini 4 Neon vs Router.com (Ramp Router)](https://www.anchorterminal.com/compare/gemini-4-neon-vs-ramp-router.md)\n- [Gemini 4 Neon vs Beam 1.0](https://www.anchorterminal.com/compare/gemini-4-neon-vs-reflection-beam.md)\n- [Gemini 4 Neon vs RouterPlus](https://www.anchorterminal.com/compare/gemini-4-neon-vs-routerplus.md)\n- [Gemini 4 Neon vs SambaCloud](https://www.anchorterminal.com/compare/gemini-4-neon-vs-sambanova.md)\n- [Gemini 4 Neon vs SiliconFlow](https://www.anchorterminal.com/compare/gemini-4-neon-vs-siliconflow.md)\n- [Gemini 4 Neon vs Simplismart](https://www.anchorterminal.com/compare/gemini-4-neon-vs-simplismart.md)\n- [Gemini 4 Neon vs StepFun API](https://www.anchorterminal.com/compare/gemini-4-neon-vs-stepfun-api.md)\n- [Gemini 4 Neon vs SWE-2 Flash](https://www.anchorterminal.com/compare/gemini-4-neon-vs-swe-2-flash.md)\n- [Gemini 4 Neon vs Tencent Cloud TokenHub](https://www.anchorterminal.com/compare/gemini-4-neon-vs-tencent-tokenhub.md)\n- [Gemini 4 Neon vs Together AI Inference](https://www.anchorterminal.com/compare/gemini-4-neon-vs-together-inference.md)\n- [Gemini 4 Neon vs Vercel AI Gateway](https://www.anchorterminal.com/compare/gemini-4-neon-vs-vercel-ai-gateway.md)\n- [Gemini 4 Neon vs Vertex AI (Gemini Enterprise Agent Platform)](https://www.anchorterminal.com/compare/gemini-4-neon-vs-vertex-ai.md)\n- [Gemini 4 Neon vs xAI API (Grok)](https://www.anchorterminal.com/compare/gemini-4-neon-vs-xai-api.md)\n- [Gemini Developer API vs Jais 3](https://www.anchorterminal.com/compare/gemini-api-vs-jais-3.md)\n- [GPT-6.2 Astra vs Jais 3](https://www.anchorterminal.com/compare/gpt-6-2-astra-vs-jais-3.md)\n- [GroqCloud vs Jais 3](https://www.anchorterminal.com/compare/groq-vs-jais-3.md)\n- [Hugging Face Inference Providers vs Jais 3](https://www.anchorterminal.com/compare/hugging-face-inference-providers-vs-jais-3.md)\n- [Tencent Hy4 vs Jais 3](https://www.anchorterminal.com/compare/hy4-vs-jais-3.md)\n- [Jais 3 vs Meta Model API](https://www.anchorterminal.com/compare/jais-3-vs-meta-model-api.md)\n- [Jais 3 vs MiniMax API](https://www.anchorterminal.com/compare/jais-3-vs-minimax-api.md)\n- [Jais 3 vs Mistral AI API](https://www.anchorterminal.com/compare/jais-3-vs-mistral-api.md)\n- [Jais 3 vs Muse Ember](https://www.anchorterminal.com/compare/jais-3-vs-muse-ember.md)\n- [Jais 3 vs Muse Spark Open](https://www.anchorterminal.com/compare/jais-3-vs-muse-spark-open.md)\n- [Jais 3 vs Nemotron 3.5 Super VL](https://www.anchorterminal.com/compare/jais-3-vs-nemotron-3-5-super-vl.md)\n- [Jais 3 vs Novita AI](https://www.anchorterminal.com/compare/jais-3-vs-novita-ai.md)\n- [Jais 3 vs OpenAI API](https://www.anchorterminal.com/compare/jais-3-vs-openai-api.md)\n- [Jais 3 vs OpenRouter](https://www.anchorterminal.com/compare/jais-3-vs-openrouter.md)\n- [Jais 3 vs Prism Inference](https://www.anchorterminal.com/compare/jais-3-vs-prism-inference.md)\n- [Jais 3 vs Router.com (Ramp Router)](https://www.anchorterminal.com/compare/jais-3-vs-ramp-router.md)\n- [Jais 3 vs Beam 1.0](https://www.anchorterminal.com/compare/jais-3-vs-reflection-beam.md)\n- [Jais 3 vs RouterPlus](https://www.anchorterminal.com/compare/jais-3-vs-routerplus.md)\n- [Jais 3 vs SambaCloud](https://www.anchorterminal.com/compare/jais-3-vs-sambanova.md)\n- [Jais 3 vs SiliconFlow](https://www.anchorterminal.com/compare/jais-3-vs-siliconflow.md)\n- [Jais 3 vs Simplismart](https://www.anchorterminal.com/compare/jais-3-vs-simplismart.md)\n- [Jais 3 vs StepFun API](https://www.anchorterminal.com/compare/jais-3-vs-stepfun-api.md)\n- [Jais 3 vs SWE-2 Flash](https://www.anchorterminal.com/compare/jais-3-vs-swe-2-flash.md)\n- [Jais 3 vs Tencent Cloud TokenHub](https://www.anchorterminal.com/compare/jais-3-vs-tencent-tokenhub.md)\n- [Jais 3 vs Together AI Inference](https://www.anchorterminal.com/compare/jais-3-vs-together-inference.md)\n- [Jais 3 vs Vercel AI Gateway](https://www.anchorterminal.com/compare/jais-3-vs-vercel-ai-gateway.md)\n- [Jais 3 vs Vertex AI (Gemini Enterprise Agent Platform)](https://www.anchorterminal.com/compare/jais-3-vs-vertex-ai.md)\n- [Jais 3 vs xAI API (Grok)](https://www.anchorterminal.com/compare/jais-3-vs-xai-api.md)\n",
  "meta": {
    "attribution": "Anchor Terminal (https://www.anchorterminal.com)",
    "attributionRequired": true,
    "credit": "CC BY 4.0, credit Anchor Terminal and link the page or API address you took it from (https://www.anchorterminal.com/terms/#data)",
    "docs": "https://www.anchorterminal.com/docs/",
    "generatedAt": "2026-11-19",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "1.0",
    "openapi": "https://www.anchorterminal.com/openapi.json",
    "preview": false,
    "run": "2026-10-01",
    "runLabel": "October 2026 research run",
    "terms": "https://www.anchorterminal.com/terms/#data"
  },
  "page": {
    "breadcrumbs": [
      {
        "name": "Home",
        "url": "https://www.anchorterminal.com/"
      },
      {
        "name": "Compare",
        "url": "https://www.anchorterminal.com/compare/"
      },
      {
        "name": "Gemini 4 Neon vs Jais 3",
        "url": ""
      }
    ],
    "description": "Gemini 4 Neon scores 65 (B) to Jais 3's 55.9 (C) for llm inference. Prices, MCP, x402, uptime and agent notes side by side.",
    "facts": [
      "Gemini 4 Neon B 65",
      "Jais 3 C 55.9",
      "scores"
    ],
    "h1": "Gemini 4 Neon vs Jais 3",
    "image": "https://www.anchorterminal.com/assets/og/default.png",
    "path": "/compare/gemini-4-neon-vs-jais-3",
    "published": "2026-10-01",
    "section": "tools",
    "title": "Gemini 4 Neon vs Jais 3 for AI agents in 2026: scores and prices",
    "toc": null,
    "updated": "2026-11-18",
    "url": "https://www.anchorterminal.com/compare/gemini-4-neon-vs-jais-3"
  },
  "tokens": {
    "markdown": 7000,
    "slim": 1030
  },
  "version": 1
}
