What It Does
Models.dev is an open-source, community-contributed database of AI model specifications. It aggregates metadata (pricing, context window sizes, capabilities, features) for models across 75+ providers including Anthropic, OpenAI, Google, Mistral, DeepSeek, xAI, and others into a structured, queryable format. Data is stored as TOML files organized by provider and model, which are used to generate a website and power a public JSON API at https://models.dev/api.json.
Built and maintained by Anomaly Innovations (the OpenCode team), Models.dev serves as the model discovery layer for OpenCode but is designed to be used independently by any application. Model IDs align with the Vercel AI SDK format for straightforward integration.
Key Features
- Open-source model metadata database covering 75+ LLM providers
- Public JSON API for programmatic access (
https://models.dev/api.json) - TOML-based data files organized by provider, enabling community contributions
- Pricing, context limits, capabilities, and feature flags per model
- Model ID format compatible with Vercel AI SDK
- Community-contributed data with pull request workflow
Use Cases
- AI application development: Querying available models and their capabilities programmatically when building multi-model applications
- Cost optimization: Comparing pricing across providers for similar capability tiers
- OpenCode integration: Used internally by OpenCode to populate provider and model selection menus
- Model selection tooling: Building internal dashboards or selection tools that need structured model metadata
Adoption Level Analysis
Small teams (<20 engineers): Easy to use — just hit the API or browse the website. No operational overhead. Useful for anyone building LLM-powered applications who needs model metadata.
Medium orgs (20-200 engineers): The API is useful for internal tooling and cost tracking. However, data accuracy depends on community contributions and may lag behind provider announcements. No SLA on the API.
Enterprise (200+ engineers): Useful as a reference but should not be the sole source of truth for production model routing. Data freshness and accuracy are community-dependent. No enterprise support or guarantees.
Alternatives
| Alternative | Key Difference | Prefer when… |
|---|---|---|
| LiteLLM Provider Registry | Part of LiteLLM proxy, more tightly integrated with routing | You use LiteLLM for model routing and want integrated metadata |
| Provider APIs directly | First-party, always up-to-date | You need guaranteed accuracy and use a single provider |
| Artificial Analysis | Independent benchmarking + metadata | You want performance benchmarks alongside metadata |
Evidence & Sources
- Models.dev GitHub Repository — source code and data files
- DeepWiki: OpenCode Provider and Model Configuration — documentation on how OpenCode uses Models.dev
- EveryDev: Models.dev Overview — independent tool listing
Notes & Caveats
- Maintained by a single company. While open-source and community-contributed, Anomaly Innovations is the primary maintainer. If the company pivots or reduces investment, the registry could become stale.
- Data accuracy is community-dependent. Provider pricing and capabilities change frequently. There is no automated verification that the TOML data matches current provider offerings.
- No SLA on the public API. The JSON API is free and public but has no uptime or latency guarantees. Production applications should cache responses.
- Tight coupling with OpenCode ecosystem. While usable independently, the project’s primary purpose is to serve OpenCode. Development priorities will reflect OpenCode’s needs.