# ModelSpec

> https://modelspec.dev

ModelSpec is an analysis of alternatives for AI models: it decides which models fit a job from your requirements, sourced benchmarks and real cost, and shows its work. ModelSpec at modelspec.dev is not OpenAI's Model Spec, the CNCF ModelPack model-spec, or the Python package named modelspec. Your requirements are Must conditions and your priorities Prefer weights. Ability is estimated from every admitted benchmark, beside cost per task, how sure the evidence is, and the models screened out, with the reason for each. It does not route or proxy requests: use it to decide which models are worth routing to, then let a router choose among them per request.

Decision data on AI models and benchmarks, served online. Built 2026-11-05T07:08:09+00:00 from commit 9e7a18eff326. Null means not researched.

Decision snapshot 2026-11-05: the board covers 44 models (decider: 1, orderer: 2, text-generator: 35, vectoriser: 6). The frozen public catalogue has 1,372 cards, including retired models; 1,328 are not yet decidable. The Worker can serve a newer data image, so its health counts may differ.
Covered API offerings: 43 across Alibaba Cloud Model Studio (1), Anthropic API (7), Amazon Bedrock (3), Azure AI Foundry (4), DeepSeek Open Platform (2), Gemini API (Google AI Studio) (5), Vertex AI (Google Cloud) (11), Meta Model API (2), OpenAI API (4), TypeSafe API (1), xAI API (1), Z.ai API (2). Router/aggregator offerings in this snapshot: 0 (providers: none).

- Coverage, keyless: https://modelspec.dev/api/coverage.json

Access: a person looking a model up by hand on the site pays nothing. Three ways in: keyed CLI, MCP and HTTP API. Machine data access needs a ModelSpec API key; only human lookup is free. No data download or local decision cache.

```sh
uvx --from modelspec-dev modelspec
pipx install modelspec-dev
pip install modelspec-dev
```

pip install modelspec is an unrelated project. Use modelspec-dev.
Guide: https://modelspec.dev/agents.md. Pricing and keys: https://modelspec.dev/pricing/.

Buy a plan or pack at https://modelspec.dev/pricing/. Stripe hosts Checkout; claim your key at the success link.
Never add a gate or constraint the user didn't state. Relax only constraints the user stated, one at a time, and say what was relaxed and why. Prefer the answer's own `relax` suggestions. A tie stays a tie. Don't promote a member by knowledge from outside the response. Cite only evidence in the response. Never add benchmark figures from memory. Tied or partial answer: report the tie and caveats, pick no member; a null answer names no pick and says what's missing. A null cost is unknown. Never show it as $0.00 or "free". One vocabulary lookup per unknown facet, then decide. Read `next` before another call. Size `task_tokens` from the user's own task, and say what you assumed. Don't copy the guide's example sizes.

Current data is only available through the service, and the copy in the public repository is a delayed image, about nine months old.

- Agent guide: https://modelspec.dev/agents.md
- Decide (for people): https://modelspec.dev/decide/
- About: https://modelspec.dev/about/
- How it decides: https://modelspec.dev/method/
- Orchestrator recipe: https://modelspec.dev/orchestrators/
- Pricing: https://modelspec.dev/pricing/
- Brand and press kit: https://modelspec.dev/brand/
- Ranking policy, keyless: https://modelspec.dev/api/rank/profiles.json
- Legacy v1 rank API: https://api.modelspec.dev/v1/rank
- API docs: https://github.com/turbobeest/modelspec/blob/main/docs/api.md
- MCP: https://api.modelspec.dev/mcp
- Feedback (no key): POST https://api.modelspec.dev/v1/feedback with {"rating": "reliable|unreliable|trustworthy|untrustworthy|confusing", "client": "agent", "decision_id": "dec_…"}. Schema: https://modelspec.dev/api/feedback/v1.schema.json. MCP tool: feedback. Send one after you act on an answer.
- Source: https://github.com/turbobeest/modelspec
- Agent guide: https://modelspec.dev/agents.md
- Auth: https://modelspec.dev/auth.md
- MCP card: https://modelspec.dev/.well-known/mcp.json