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 Legacy models that are still available but superseded live in their own table at the bottom.
| Model | API ID | Context | Max output | Best for |
|---|---|---|---|---|
| Claude Opus 4.8 | claude-opus-4-8 | 1M tokens | 128k | Most complex reasoning + agentic coding; default for hardest tasks |
| Claude Sonnet 4.6 | claude-sonnet-4-6 | 1M tokens | 64k | Best speed/quality balance; default for most agent workloads |
| Claude Haiku 4.5 | claude-haiku-4-5-20251001 | 200k tokens | 64k | Cheap, fast — classification, routing, high-volume tasks |
Claude Sonnet 4.6 + Haiku 4.5 support extended thinking; Opus 4.8 uses adaptive thinking (always on) and does not offer extended thinking. All support vision, 1M tokens on Opus/Sonnet, and the Message Batches API for 50% off. On Opus 4.8 the temperature / top_p / top_k params are deprecated (a non-default value returns a 400) — steer behavior through prompting instead.
Aliases (claude-opus-4-8, claude-sonnet-4-6, claude-haiku-4-5) auto-resolve to the latest snapshot. From the 4.6 generation onward the dateless IDs are themselves pinned snapshots, not evergreen pointers; pin dated IDs (claude-haiku-4-5-20251001) for reproducibility.
A higher tier, Claude Fable 5 (claude-fable-5, 1M context, 128k output, $10/$50 per MTok), is generally available as of June 2026 for the most demanding reasoning + long-horizon agentic work.
Source: platform.claude.com/docs/en/about-claude/models/overview.
OpenAI's model lineup moves fastest. As of 2026-06-12, the docs index lives at developers.openai.com/api/docs/models (the old platform.openai.com/docs/models now 301-redirects there) — visit it for the authoritative current list. The current frontier model is GPT-5.5 (gpt-5.5, 1M context), with GPT-5.4 (gpt-5.4) and GPT-5.4 mini (gpt-5.4-mini, 400k context) below it. In general, expect:
| Model | Best for |
|---|---|
| Gemini 3.1 Pro | Complex reasoning + coding; flagship |
| Gemini 3.5 Flash | Most intelligent Flash; 1M context |
| Gemini 3 Flash | Fast general-purpose (preview) |
| Gemini 3.1 Flash-Lite | Highest throughput |
| Gemini 2.5 Pro / Flash | Legacy, still available |
All Gemini 3 models carry a 1M-token context window. Source: ai.google.dev/gemini-api/docs/models.
| Model family | Provider | Notes |
|---|---|---|
| Llama (code variants) | Meta | Run locally via Ollama / vLLM; varies by size |
| Codestral | Mistral | Open-source, code-focused |
| DeepSeek | DeepSeek | Strong SWE-bench performance, low cost |
| Qwen Coder | Alibaba | Code-focused, multiple sizes |
Run locally via Ollama, vLLM, or llama.cpp. Latency and capability ceiling are lower than frontier models, but data stays on your machine.
These models continue to work but are superseded — consider migrating:
| Model | API ID | Status |
|---|---|---|
| Claude Opus 4.7 | claude-opus-4-7 | Superseded by Opus 4.8 |
| Claude Opus 4.6 | claude-opus-4-6 | Superseded by Opus 4.8 |
| Claude Sonnet 4.5 | claude-sonnet-4-5-20250929 | Superseded by Sonnet 4.6 |
| Claude Opus 4.5 | claude-opus-4-5-20251101 | Legacy |
| Claude Opus 4.1 | claude-opus-4-1-20250805 | Deprecated, retires 2026-08-05 |
| Claude Sonnet 4 | claude-sonnet-4-20250514 | Deprecated, retires 2026-06-15 |
| Claude Opus 4 | claude-opus-4-20250514 | Deprecated, retires 2026-06-15 |
| Claude Haiku 3 | claude-3-haiku-20240307 | Retired 2026-04-20 |
Always check model deprecations before pinning an older ID.
Rule of thumb, optimized for cost/quality:
See Cost Optimization § Model Routing for the split-model pattern.