{"service":"modelintel","summary":"Deterministic fusion of open LLM/AI-model catalogs — per-1M pricing, context/capability metadata and real Aider coding-benchmark scores — with a router that returns the cheapest model passing a task's constraints. x402-native, keyless, NO LLM in the hot path.","tos_note":"Sources are permissively-licensed open catalogs ONLY. The OpenRouter API is NOT used: its ToS forbids scraping its Site/Services and developing a competing service.","sources":[{"key":"litellm","name":"LiteLLM model_prices_and_context_window.json","kind":"pricing","url":"https://raw.githubusercontent.com/BerriAI/litellm/main/model_prices_and_context_window.json","license":"MIT","attribution":"LiteLLM (BerriAI) — github.com/BerriAI/litellm, MIT license.","note":"Per-token USD pricing (x1e6 -> per-1M), context/output limits, provider, and ~40 capability flags. The broadest keyless price catalog.","entry_count":2415,"status":"ok"},{"key":"modelsdev","name":"models.dev api.json","kind":"metadata","url":"https://models.dev/api.json","license":"MIT","attribution":"models.dev — github.com/sst/models.dev, MIT license.","note":"Per-1M USD pricing, context/output limits, modalities, knowledge cutoff, release date, reasoning/tool_call/attachment flags per provider.","entry_count":5756,"status":"ok"},{"key":"aider_polyglot","name":"Aider polyglot coding leaderboard","kind":"benchmark","url":"https://raw.githubusercontent.com/Aider-AI/aider/main/aider/website/_data/polyglot_leaderboard.yml","license":"Apache-2.0","attribution":"Aider (Aider-AI) — github.com/Aider-AI/aider, Apache-2.0 license.","note":"Real coding-benchmark pass rates (pass_rate_2 %) on the polyglot suite; the primary coding-quality anchor for the router.","entry_count":39,"status":"ok"},{"key":"aider_edit","name":"Aider edit-format coding leaderboard","kind":"benchmark","url":"https://raw.githubusercontent.com/Aider-AI/aider/main/aider/website/_data/edit_leaderboard.yml","license":"Apache-2.0","attribution":"Aider (Aider-AI) — github.com/Aider-AI/aider, Apache-2.0 license.","note":"Older edit-format coding pass rates; a secondary benchmark anchor.","entry_count":47,"status":"ok"}],"tasks":["coding","reasoning","vision","tool-use","long-context","cheap-chat","general"],"capability_tags":["tools","parallel_tools","vision","reasoning","pdf","audio","web_search","prompt_caching","structured_output"],"index":{"models_indexed":3775,"built_ts":1785114834,"ttl_seconds":21600,"sources_used":["litellm","modelsdev","aider_polyglot","aider_edit"],"degraded":[]},"methodology":"Fetch three fixed keyless sources -> normalize ids to a canonical slug -> fuse by slug (Aider matched by fuzzy model name) -> prices normalized to per-1M USD (LiteLLM per-token x1e6, models.dev already per-1M). On a price divergence BOTH values are kept, each with its own source citation (no invented truth). Unknown fields are honest null.","moat":"Provider-neutral (all catalog providers, not one routable catalog), fused pricing + capabilities + real coding benchmarks, plus a deterministic cheapest-passing-constraints router — x402-native, no account.","not_advice":"Verify pricing/capabilities against each provider before production routing; catalogs can lag provider changes."}