AI Model Feature Support Matrix

This is not a calculator — it is a capability lookup, so correctness means one thing: fidelity to the official provider documentation. The matrix covers 28 current models across seven providers (OpenAI, Anthropic, Google, Meta, Mistral, DeepSeek and xAI) and 14 developer features — 392 cells in total, each set by hand from a provider doc rather than guessed or scraped.

Every cell resolves to one of three states:

• ✓ Supported— the feature is generally available on that model's hosted API.
• ~ Partial — real but caveated: in beta or preview, delivered through a separate tool or endpoint, or model-specific. Every partial cell carries a note in the detail panel explaining exactly why.
• — Not supported— the provider docs do not offer it on that model's API.

The filtering is a set of pure, deterministic functions. Selecting feature chips runs matchesFeatures(model, selected, mode): in Match ALL mode a model qualifies only when every selected feature is supported or partial; in Match ANY mode it qualifies when at least one is. Provider and model-name filters compose on top with plain set membership and a case-insensitive substring match. An empty selection is a no-op, so the matrix always shows the full catalogue by default rather than a dead, empty page.

For credibility the module also exposes a compareModels(a, b) cross-check: it reads the exact same feature records the matrix renders and returns a feature-by-feature diff, so a single cell and the compare panel can never disagree. A matrixCoverage() self-test confirms all 392 cells resolve to a concrete value — no silent gaps. Because everything is static, there are no live API calls; the table is versioned with a LAST_VERIFIED date (2026-07-17) and every model row links to the provider doc it was read from.

Each model row in the tool links to the exact provider doc its values came from. The primary references are:

• OpenAI — Models & API reference
• Anthropic — Claude models overview
• Anthropic — Prompt caching & Message Batches API
• Google — Gemini API models
• Meta — Llama models overview
• Mistral — models overview
• DeepSeek — API documentation
• xAI — Grok models

Feature values were last cross-checked against these sources on 2026-07-17. Model capabilities change frequently — always confirm against the linked provider doc before writing production code.