Spec updates

Changelog — what we've updated

We curate the catalog manually — every spec is regularly cross-checked against official provider documentation. This page shows what was reviewed and when.

May 2026

  1. anthropic/claude-haiku-4.5
    2026-05-25

    Fastest Claude — classification, routing, simple agents at mini-tier prices.

  2. anthropic/claude-opus-4.6
    2026-05-25

    Anthropic's premium tier for the hardest reasoning and coding tasks.

  3. anthropic/claude-sonnet-4.6
    2026-05-25

    Anthropic's workhorse: top-tier coding and agentic, fair pricing.

  4. deepseek/deepseek-v3
    2026-05-25

    DeepSeek-V3 — MoE model, impressive coding quality for very little money.

  5. google/gemini-3-flash
    2026-05-25

    Fast, cheap Gemini — 1M context, fully multimodal.

  6. google/gemini-3-pro
    2026-05-25

    Google's flagship: 2M token context, native video and audio input.

  7. google/gemini-3.1-flash-image-preview
    2026-05-25

    Google's multimodal image-generation model — image output straight from the prompt.

  8. meta/llama-3.3-70b
    2026-05-25

    Meta's open-weight workhorse — cheap and solid for standard tasks.

  9. mistral/mistral-large-2
    2026-05-25

    Mistral's flagship — EU-hosted, strong on multilingual generation.

  10. openai/gpt-5.4
    2026-05-25

    OpenAI's new flagship — agentic workloads and deep reasoning chains.

  11. openai/gpt-5.4-mini
    2026-05-25

    GPT-5.4's smaller sibling — cheaper, faster, built for volume.

  12. openai/gpt-image-1
    2026-05-25

    OpenAI's current image-generation model — precise prompt following, text-in-image.

  13. openai/text-embedding-3-large
    2026-05-25

    OpenAI's large embedding model — 3072-dim, very strong semantic search.

  14. openai/text-embedding-3-small
    2026-05-25

    OpenAI's small embedding model — 1536-dim, very cheap, broadly applicable.

  15. xai/grok-4
    2026-05-25

    xAI's flagship with live X (Twitter) data access and a reasoning mode.

A date means: on that day we verified the spec against provider documentation and updated it as needed.