Models (Unified)

All model IDs and context windows verified against vendor docs on the lastVerified date above. Legacy models that are still available but superseded live in their own table at the bottom.

Anthropic (current)

All three support extended thinking (Sonnet + Haiku) or adaptive thinking (Opus), vision, 1M tokens on Opus/Sonnet, and the Message Batches API for 50% off.

Aliases (claude-opus-4-7, claude-sonnet-4-6, claude-haiku-4-5) auto-resolve to the latest snapshot. Pin dated IDs (claude-haiku-4-5-20251001) for reproducibility.

Source: platform.claude.com/docs/en/about-claude/models/overview.

OpenAI

OpenAI's model lineup moves fastest. As of 2026-04-17, visit platform.openai.com/docs/models for the authoritative current list — our verification fetch on this page returned 403 during the audit. In general, expect:

• A flagship reasoning model (most capable, highest cost, slowest)
• A mid-tier general-purpose chat model
• A small/nano tier for classification + routing
• A realtime model family for voice / streaming use cases

Google (Gemini)

Source: ai.google.dev/gemini-api/docs/models.

Open-source / local

Run locally via Ollama, vLLM, or llama.cpp. Latency and capability ceiling are lower than frontier models, but data stays on your machine.

Anthropic legacy (still available)

These models continue to work but are superseded — consider migrating:

Always check model deprecations before pinning an older ID.

Picking a model for a coding agent

Rule of thumb, optimized for cost/quality:

• Default workhorse → Claude Sonnet 4.6 (agentic coding sweet spot) or OpenAI current flagship
• Hardest tasks / long-horizon agents → Claude Opus 4.7
• Classification, routing, summarization → Claude Haiku 4.5 or Gemini 3 Flash
• Local / privacy-first → DeepSeek Coder or Qwen Coder via Ollama

See Cost Optimization § Model Routing for the split-model pattern.

Further reading

• Pricing — what these models actually cost
• Landscape Overview — which tools use which models
• Anthropic model migration guide