{
  "listings": [
    {
      "slug": "celeris-1-decision",
      "name": "Celeris-1 Decision",
      "company": "Marqo Inc",
      "grade": "D",
      "fit": "yes",
      "why": "noul returns the probability of yes; the Policy gates section says to ask whether the input violates the policy and pick thresholds from labelled examples. 131,072-token context, so a long policy fits.",
      "source": "https://docs.celeris.ai/decisions.md",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/celeris-1-decision"
    },
    {
      "slug": "cloudflare-clef",
      "name": "Clef",
      "company": "Cloudflare, Inc.",
      "grade": "B",
      "fit": "yes",
      "why": "noul returns the probability of yes; 65,536-token context on Clef. Arize found Clef kept 0.92 and 0.91 AUC with the question worded both ways, one of three models that did. No prompt-injection guidance from Cloudflare.",
      "source": "https://developers.cloudflare.com/workers-ai/models/clef/",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/cloudflare-clef"
    },
    {
      "slug": "elephant-1",
      "name": "elephant-1",
      "company": "Baseten Labs, Inc.",
      "grade": "D",
      "fit": "yes",
      "why": "noul returns P(yes); trained at 16,384 tokens with longer requests refused, a trained 'I don't know' option and 16 percent safety and abstention data. No prompt-injection guidance.",
      "source": "https://huggingface.co/baselabs/elephant-1",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/elephant-1"
    },
    {
      "slug": "joey-1",
      "name": "joey-1",
      "company": "Baseten Labs, Inc.",
      "grade": "C",
      "fit": "yes",
      "why": "noul returns P(yes); trained at 16,384 tokens with longer requests refused, a trained 'I don't know' option and 16 percent safety and abstention data. No prompt-injection guidance.",
      "source": "https://huggingface.co/baselabs/joey-1",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/joey-1"
    },
    {
      "slug": "mammoth-1",
      "name": "mammoth-1",
      "company": "Baseten",
      "grade": "D",
      "fit": "yes",
      "why": "noul returns P(yes); trained at 16,384 tokens with longer requests refused, a trained 'I don't know' option and 16 percent safety and abstention data. No prompt-injection guidance.",
      "source": "https://huggingface.co/baselabs/mammoth-1",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/mammoth-1"
    },
    {
      "slug": "nace-drex",
      "name": "Drex 1.5",
      "company": "Nace AI, Inc.",
      "grade": "BB",
      "fit": "yes",
      "why": "noul with true and false criteria; drex-v2.0 reads states to 131,072 tokens and refuses longer ones with a 422 naming the limit, so a long policy need not be compressed.",
      "source": "https://console.nace.ai/docs/drex/guides/questions",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/nace-drex"
    },
    {
      "slug": "openai-decisions-api",
      "name": "OpenAI Decisions API",
      "company": "OpenAI OpCo, LLC",
      "grade": "A",
      "fit": "yes",
      "why": "predicate returns P(true); input can carry conversation history, tool calls and tool outputs; thresholds from labelled examples; 922,000 input tokens on GPT-6 Luna. Arize found it kept 0.90 and 0.89 AUC with the question reversed. No adversarial-input guidance.",
      "source": "https://developers.openai.com/api/docs/guides/decisions",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/openai-decisions-api"
    },
    {
      "slug": "pplx-decider",
      "name": "pplx-decider",
      "company": "Perplexity AI, Inc.",
      "grade": "BB",
      "fit": "yes",
      "why": "noul with criteria, 256K context, and published calibration (ECE 0.014, Brier 0.134 on the development set; about 0.05 on held-out decision datasets). No injection guidance.",
      "source": "https://docs.perplexity.ai/docs/decisions/quickstart",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/pplx-decider"
    },
    {
      "slug": "pplx-decider-flash",
      "name": "pplx-decider-flash",
      "company": "Perplexity AI, Inc.",
      "grade": "B",
      "fit": "yes",
      "why": "noul with criteria, 256K context, calibration published (ECE 0.021, Brier 0.163 on the development set).",
      "source": "https://docs.perplexity.ai/docs/decisions/quickstart",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/pplx-decider-flash"
    },
    {
      "slug": "sqwish-kettei",
      "name": "kettei",
      "company": "Sqwish Ltd, trading as Sqwish Labs",
      "grade": "C",
      "fit": "yes",
      "why": "binary decisions with abstain_below (examples at 0.4 to 0.7) and a tool-call check example; 65,536-token input limit with a 422 context_too_long rather than a cut.",
      "source": "https://docs.sqwish.ai/llms-full.txt",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/sqwish-kettei"
    },
    {
      "slug": "typesafe-jev",
      "name": "Jev",
      "company": "TypeSafe AI, Inc.",
      "grade": "B",
      "fit": "yes",
      "why": "noul with 0.7 and 0.3 as starting thresholds and a worked risk-tiered example (0.5 floor, 0.9 for a transfer); OpenRouter's screen caught 40 of 40 injections; the jaggedness page says adversarial content can steer answers; 64k context, 262k by header.",
      "source": "https://docs.typesafe.ai/confidence",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/typesafe-jev"
    },
    {
      "slug": "decider",
      "name": "Decider",
      "company": "Mark Marosi (Mapika)",
      "grade": "BB",
      "fit": "caveat",
      "why": "noul with per-type temperatures, but the README says rules written into the question are not followed at this size (0.24 against 0.67) and calibration on hard items is weak; the maintainer advised acting on destructive actions only at 0.8 or above.",
      "source": "https://github.com/Mapika/decider#readme",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/decider"
    },
    {
      "slug": "decision-3",
      "name": "Decision 3.0",
      "company": "vLLM Semantic Router project",
      "grade": "B",
      "fit": "caveat",
      "why": "noul with a probability of yes and no truncation, but issue 6298 found the vela3 and d3 sizes used as jailbreak and harm guards missed attacks the base model refused.",
      "source": "https://huggingface.co/vllm-sr/d3",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/decision-3"
    },
    {
      "slug": "drex-max-v1",
      "name": "Drex Max v1",
      "company": "Nace AI, Inc.",
      "grade": "C",
      "fit": "caveat",
      "why": "noul with a 16,384-token default raisable to 262,144 (DREX_MAX_CONTEXT). No safety results, no prompt-injection guidance and the licence cap.",
      "source": "https://huggingface.co/nace-ai/drex-max-v1",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/drex-max-v1"
    },
    {
      "slug": "granite-switch",
      "name": "Granite Switch 4.1",
      "company": "International Business Machines Corporation",
      "grade": "B",
      "fit": "caveat",
      "why": "A guardian-core safety adapter and a decision adapter share the endpoint, with the 128K base context for a long policy; there is no typed noul, so the gate is a two-label logprob read.",
      "source": "https://huggingface.co/ibm-granite/granite-switch-4.1-8b",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/granite-switch"
    },
    {
      "slug": "jaredpalmer-kev",
      "name": "Kev",
      "company": "Jared Palmer",
      "grade": "BB",
      "fit": "caveat",
      "why": "noul with calibrated probabilities and a 65,536-token server limit, but Arize found Kev 9B's AUC fall from 0.85 to 0.54 when the question was reversed, and the small models are validated to 8,192 tokens.",
      "source": "https://github.com/jaredpalmer/kev#readme",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/jaredpalmer-kev"
    },
    {
      "slug": "jev-27b-vl",
      "name": "JEV-27B-VL",
      "company": "AutoTrust AI Lab",
      "grade": "C",
      "fit": "caveat",
      "why": "Yes/no decisions over up to 256K tokens of text (20 of 20 at every depth tested), one question a request, with no safety results or injection guidance.",
      "source": "https://huggingface.co/autotrust/JEV-27B-VL",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/jev-27b-vl"
    },
    {
      "slug": "liquid-d1",
      "name": "Liquid d1",
      "company": "Liquid AI, Inc.",
      "grade": "C",
      "fit": "caveat",
      "why": "noul with a worked binary-decision example in the guide, but Arize saw d1's AUC fall from 0.92 to 0.17 when the question was reversed, so the gate must be tested both ways.",
      "source": "https://docs.liquid.ai/guides/decision-model-guide",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/liquid-d1"
    },
    {
      "slug": "microsoft-decision-1",
      "name": "Microsoft-Decision-1",
      "company": "Microsoft",
      "grade": "E",
      "fit": "caveat",
      "why": "noul with criteria for true and false, 32,768-token context and a vendor safety test over 5,250 requests; no refusal type, and 0.42 s on OpenRouter is over the 300 ms target.",
      "source": "https://learn.microsoft.com/en-us/azure/foundry/foundry-models/how-to/use-foundry-models-microsoft-decision",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/microsoft-decision-1"
    },
    {
      "slug": "rightnow-system-zero",
      "name": "RightNow System Zero",
      "company": "RightNow AI, Inc.",
      "grade": "B",
      "fit": "caveat",
      "why": "boolean returns the probability of true with a refusal per question, over 38,000 state tokens, but the probabilities are flagged uncalibrated, so a threshold has no stated meaning.",
      "source": "https://www.rightnowai.co/docs/system-zero/decide",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/rightnow-system-zero"
    },
    {
      "slug": "scaledown",
      "name": "ScaleDown",
      "company": "Carbon ScaleDown, Inc.",
      "grade": "E",
      "fit": "caveat",
      "why": "noul returns an independent P(yes); no context limit, calibration figure or refusal type is documented, and no key without sales contact.",
      "source": "https://docs.scaledown.ai/api-reference/decisions",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/scaledown"
    },
    {
      "slug": "sprag-spev",
      "name": "SPEV",
      "company": "Sprag AI, Inc.",
      "grade": "D",
      "fit": "caveat",
      "why": "predicate returns P(true) with a refusal type and a 400 model_limit_exceeded naming the limit; the weights' config caps input at 8,192 tokens, short for a long fixed policy.",
      "source": "https://sprag.ai/docs/decisions/api",
      "confidence": "medium",
      "url": "https://www.anchorterminal.com/tools/sprag-spev"
    },
    {
      "slug": "strands-decider",
      "name": "Strands Decider 2B",
      "company": "Amazon Web Services, Inc.",
      "grade": "B",
      "fit": "caveat",
      "why": "noul works, but at a 0.95 threshold no booking proceeded and at 0.50 two unauthorised ones passed; injection recall 92.5 percent at 17 percent false positives, weak on encoded attacks; a question naming tyres fired on every 'tire' line; the 4,096-token window blocks a 10k-token policy.",
      "source": "https://builder.aws.com/content/3K7xLOIkPXOLkH0ZZqyH2ckYjUI/after-laya-can-you-skip-training-your-own-classifier-with-strands-decider",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/strands-decider"
    },
    {
      "slug": "vela",
      "name": "Vela 2.0",
      "company": "vLLM Semantic Router project (with KR Labs)",
      "grade": "BB",
      "fit": "caveat",
      "why": "noul plus trained safety, PII and hallucination span heads with AUC over 14 safety sets, but issue 6298 found the trained adapters missed attacks the base model refused.",
      "source": "https://huggingface.co/vllm-sr/Vela-2.0-4B",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/vela"
    },
    {
      "slug": "convai-laya",
      "name": "Laya",
      "company": "Convai Innovations Pvt. Ltd.",
      "grade": "BB",
      "fit": "no",
      "why": "noul exists, but Arize scored Laya 57.3 percent on evaluator suites with calibration error 0.330 and a toxic-content question flipped from 0.11 to 0.87 AUC on rewording; an HN tester saw a die roll of 3 come back 9 percent odd at 91 percent confidence.",
      "source": "https://arize.com/blog/decision-model-benchmark.md",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/convai-laya"
    },
    {
      "slug": "gliclass",
      "name": "GLiClass",
      "company": "Knowledgator Engineering Ltd.",
      "grade": "C",
      "fit": "no",
      "why": "No predicate type with a calibrated P(yes); labels get softmax or independent sigmoid scores, F1 is the only published metric and issue 12 shows saturation at 0.99.",
      "source": "https://docs.knowledgator.com/docs/frameworks/gliclass/usage/",
      "confidence": "high",
      "url": "https://www.anchorterminal.com/tools/gliclass"
    }
  ],
  "marks": "yes: the docs show it does the job as the success test says. caveat: it does the job with a limit an agent should know. no: it doesn't do this job. unchecked: we couldn't read it.",
  "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": "2026-11-16",
    "license": "CC-BY-4.0",
    "method": "https://www.anchorterminal.com/benchmark/",
    "methodology": "0.8",
    "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"
  },
  "useCase": {
    "accounts": 8,
    "after": [
      "sandboxes.run-command",
      "human-review",
      "agent-frameworks.refuse-tool-call"
    ],
    "area": "models",
    "asks": [
      "is this tool call safe to run",
      "does this message authorise the booking",
      "is this request a prompt injection",
      "does this line affect driving safety"
    ],
    "before": [
      "agent-frameworks.propose-tool-call",
      "chat.receive-message"
    ],
    "blockers": [
      "disguised or encoded attacks pass a text-only gate",
      "threshold at 0.9 blocks every legitimate action, at 0.5 lets unauthorised ones through",
      "4,096-token window too short for a long fixed policy"
    ],
    "capabilities": [
      "inference.decision",
      "guard.injection"
    ],
    "category": "decision-models",
    "categoryName": "Decision models",
    "checked": "2026-11-12",
    "companies": {
      "fit": 9,
      "fitWithLimit": 12,
      "tagged": 42
    },
    "competition": "crowded",
    "fits": [
      "celeris-1-decision",
      "cloudflare-clef",
      "elephant-1",
      "joey-1",
      "mammoth-1",
      "nace-drex",
      "openai-decisions-api",
      "pplx-decider",
      "pplx-decider-flash",
      "sqwish-kettei",
      "typesafe-jev"
    ],
    "fitsWithLimit": [
      "decider",
      "decision-3",
      "drex-max-v1",
      "granite-switch",
      "jaredpalmer-kev",
      "jev-27b-vl",
      "liquid-d1",
      "microsoft-decision-1",
      "rightnow-system-zero",
      "scaledown",
      "sprag-spev",
      "strands-decider",
      "vela"
    ],
    "id": "decision-models.gate-action",
    "listings": {
      "caveat": 13,
      "no": 2,
      "notRead": 1,
      "tagged": 48,
      "unchecked": 0,
      "yes": 11
    },
    "name": "Yes or no gate before an action runs",
    "status": "draft",
    "success": "A probability of yes that can be thresholded. Measured: at the chosen threshold no irreversible action proceeds on a wrong yes in the caller's labelled set, under 300 ms.",
    "url": "https://www.anchorterminal.com/api/v1/use-cases/decision-models.gate-action.json",
    "weight": 16,
    "weightBasis": "8 of 41 accounts; the AWS nine-desk test, two GitHub issues on wrong yeses, the Vela guard issue and the Dataiku exception threshold work"
  }
}
