Model Landscape
Checked 2026-10-31. Model lineups change every few months - treat this page as a map and confirm IDs on the official pages before using them. Values marked verify could not be confirmed from an official source at the time of writing.
Vendors
| Vendor | Families | Weights | Current flagship (verified) | Official model list |
|---|---|---|---|---|
| Anthropic | Claude Fable, Opus, Sonnet, Haiku | Closed (API, Bedrock, Google Cloud, Microsoft Foundry) | Claude Opus 5.5, Sonnet 5.5, Haiku 5.5, Fable 5.1 | Models overview |
| OpenAI | GPT (incl. reasoning models) | Closed, plus open-weight releases | GPT-6 Astra (gpt-6-astra, recommended starting point); other flagship IDs gpt-6.1-sol, gpt-6.1-terra, gpt-6-luna. Open weights: gpt-oss-120b, gpt-oss-20b | Models |
| Gemini (closed), Gemma (open) | Both | Gemini 4 Argon (gemini-4-argon, GA 2026-10-13), Gemini 3.8 Flash (release notes) | Gemini models | |
| Meta | Llama | Open weights (custom license) | Llama 4 Scout, Llama 4 Maverick (newest Llama generation); Meta's newer models ship as the Muse line (e.g. Muse Spark, open-weight Muse Spark Open) on dev.meta.ai | llama.com |
| Mistral AI | Mistral, Codestral, and others | Both | Mistral Large 4 (mistral-large-4, released 2026-10-06, API public preview, open weights), Mistral Small 4 (Apache 2.0) | Mistral models |
| DeepSeek | DeepSeek V / R series | Open weights | DeepSeek-V4-Pro (deepseek-v4-pro), DeepSeek-V4.1-Flash (deepseek-flash, serves DeepSeek-V4.1-Flash-1017) | DeepSeek API docs |
| Alibaba | Qwen | Mostly open weights | Qwen3.8 (Qwen3.8-2.4T-A95B and Qwen3.8-27B, open weights, 2026-08) | Qwen on GitHub |
Anthropic Claude (verified)
| Model | API ID | Positioning (from Anthropic docs) |
|---|---|---|
| Claude Fable 5.1 | claude-fable-5-1 | Demanding reasoning and long-horizon agentic work |
| Claude Opus 5.5 | claude-opus-5-5 | Long-running agentic coding and knowledge work; recommended default |
| Claude Sonnet 5.5 | claude-sonnet-5-5 | Best combination of speed and intelligence |
| Claude Haiku 5.5 | claude-haiku-5-5 | Fastest; high-volume classification, extraction, routing |
- All four list a 1M-token context window, text + image input, tool use and adaptive thinking.
- Dateless IDs (from the 4.6 generation on) are pinned snapshots, not moving aliases.
- Older models (Claude 5, 4.x) remain available as legacy; see the deprecations page for retirement dates.
How to choose
- Start with a strong general model to see if the task is feasible.
- Build an eval set (Evals).
- Try cheaper/faster models against it; keep the cheapest one that passes.
- Need privacy/offline/self-hosting → open weights (Local Models).
- Re-run evals when vendors ship new versions.