Gemini 4 Carbon vs Jais 3

Gemini 4 Carbon 20

Model reading for gemini-4-carbon 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). Introductory prices published in Google's 28 October announcement without a login, $2 and $10, then $4 and $20; the Gemini API pricing page as read on 10 November showed no Carbon row, so the figures rest on the announcement (20). No free tier: Google's developer post puts Carbon on the paid tier (0). A Google account and AI Studio for a key (0). 0 + 20 + 0 + 0 = 20.

Jais 3 60

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.

Gemini 4 Carbon 85

Model reading for gemini-4-carbon 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 (10). The Cloud developer guide says what Carbon takes and refuses, the parameters that return 400 or are ignored, and how to migrate (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.

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.

Gemini 4 Carbon 93

Model reading for gemini-4-carbon 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 (13 of 15). Implicit caching with durable caching on by default, cached input 95% off, explicit caches not supported, per the Cloud guide read on 10 November (15). 1,048,576 tokens shared between input and output (15). Batch at half price on the platform (10). Python and JavaScript SDKs (10). Error codes with backoff guidance (10 of 15). 20 + 13 + 15 + 15 + 10 + 10 + 10 = 93.

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.

Gemini 4 Carbon 55

Model reading for gemini-4-carbon 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). Carbon is on the paid tier only, and Google's terms say paid-tier content isn't used to improve products; AI Studio use under a Google AI Ultra subscription falls under terms we didn't read, so 15 of 20 (a judgement call; the platform's 10 is for its free tier). Abuse retention with 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.

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.

Gemini 4 Carbon 83

Model reading for gemini-4-carbon 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 to developers on 10 November 2026, with the platform's changelog (30). Deprecations with earliest dates and no stated minimum notice (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.

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.

Gemini 4 Carbon 72

Model reading for gemini-4-carbon 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 (15 of 20). Data may be cached anywhere Google has facilities; no subprocessor list checked (5 of 20). Every benchmark figure is Google's own. 15 + 20 + 15 + 5 = 55.

Jais 3 55

Apache-2.0 in every card's metadata, with staged training checkpoints as branches and training logs linked on Weights & 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.