Last updated:
Claude
Anthropic's model family, built with a heavy emphasis on Constitutional AI — currently the Fable 5.1, Opus 5.5, Sonnet 5.5, and Haiku 5.5 lineup.
Model Facts
Latest model: Claude Haiku 5.5
- Provider
- Anthropic
- Released
- Sep 22, 2026
- Context window
- 1M tokens
- Weights
- Closed
Full Model List
| Model | Released | Note |
|---|---|---|
| Claude Haiku 5.5 | Oct 7, 2026 | fast, low-cost tier for high-volume work |
| Claude Sonnet 5.5 | Sep 28, 2026 | about 30% faster and up to 30% cheaper than Sonnet 5 |
| Claude Opus 5.5 | Sep 22, 2026 | matches Fable 5.1 on most work at about 40% lower cost than Opus 5 |
| Claude Fable 5.1 | Sep 1, 2026 | most capable generally available model, 1M-token context |
| Claude Mythos 5.1 | Sep 1, 2026 | limited trusted-access program only (e.g. defensive cybersecurity) |
| Claude Opus 4.5 | Nov 24, 2025 | earlier Opus generation |
| Claude Sonnet 4.5 | Sep 29, 2025 | earlier Sonnet generation |
| Claude 3.5 Sonnet | Jun 20, 2024 | prior-generation flagship |
- Official model page:
- anthropic.com/claude
- Official documentation:
- docs.anthropic.com
- Other sources:
- Opus 5.5 announcement
- Anthropic newsroom
What it is
Claude is Anthropic's model family, built with a heavy emphasis on safety and Constitutional AI. The current lineup is Claude Fable 5.1 (September 1, 2026), Opus 5.5 (September 22), Sonnet 5.5 (September 28) and Haiku 5.5 (October 7). Opus 5.5 is the entry shown here because it is Anthropic's most capable Opus model and, per Anthropic, performs at the level of Fable 5.1 on most work.
The lineup
- Fable 5.1 — the most capable generally available model, for demanding reasoning, long-running agents, coding and research; 1M-token context and up to 128K output tokens.
- Opus 5.5 — long-running agentic coding and complex knowledge work; priced at $4 / $20 per million input / output tokens.
- Sonnet 5.5 — the everyday tier for coding and document work; about 30% faster than Sonnet 5.
- Haiku 5.5 — the fastest and cheapest tier, built for high-volume, cost-sensitive work.
- Mythos 5.1 — available only through a limited trusted-access program.
Where these numbers come from
The benchmarks are Anthropic's published figures from the Opus 5.5 announcement (September 22, 2026): 66.4% on Terminal-Bench 4.0 and 81.8% on OSWorld 2.1 (computer use). Vendor-reported scores use each lab's own harness, so they are not directly comparable with other labs' numbers.
Notable traits
- A tiered family (Haiku / Sonnet / Opus / Fable) rather than a single model.
- Strong focus on agentic coding and computer use.
- Closed weights — available through the Claude apps, the Anthropic API and major cloud platforms.