The models
Four frogs on the log, plus one behind a velvet rope. Anthropic's own advice is short: start with Opus 5.5 for most work, and move to Fable 5.2 for demanding reasoning and long-horizon agentic work, or when Opus 5.5 at higher effort still falls short.
Four frogs sit on a log. One decides to jump. How many are left? Four. Deciding is not jumping. Pick a model.
The current lineup
| Fable 5.2 | Opus 5.5 | Sonnet 5.5 | Haiku 5.5 | |
|---|---|---|---|---|
| Built for | Demanding reasoning and long-horizon agentic work | Long-running agentic coding and knowledge work | The best combination of speed and intelligence | High-volume, latency-sensitive tasks such as classification, extraction and routing |
| API model ID | claude-fable-5-2 | claude-opus-5-5 | claude-sonnet-5-5 | claude-haiku-5-5 |
| Relative speed | Moderate | Moderate | Fast | Fastest |
| Input, per million tokens | $8 | $4 | $2 | From $0.10 |
| Output, per million tokens | $40 | $20 | $10 | From $0.50 |
| Context window | 1M tokens | 1M tokens | 1M tokens | 1M tokens |
| Max output | 128K tokens | 128K tokens | 128K tokens | 128K tokens |
| Default effort | high | medium | high | medium |
| Knowledge cutoff | Jun 2026 | Jun 2026 | Jun 2026 | Jun 2026 |
| Retires no sooner than | Nov 9, 2027 | Sep 22, 2027 | Sep 28, 2027 | Oct 7, 2027 |
All four take text and images in, put text out, and support tool use. A million tokens is roughly 555,000 words on the current tokenizer.
Haiku 5.5 has two price tiers: $0.10 in and $0.50 out per million tokens for prompts up to 100,000 tokens, and $0.50 in and $2.50 out above that.
Mythos 5.2
Claude Mythos 5.2 launched alongside Fable 5.2 on November 9, 2026. It has limited availability and is priced the same as Fable 5.2. If you are not sure whether you have access, you do not. Anthropic explains the access rules here.
Picking one
- Start hereOpus 5.5. This is Anthropic's stated default for most workloads.
- Step upFable 5.2 when the task is long, hard, and Opus at higher effort is not getting there. It costs twice as much per token.
- Step downSonnet 5.5 when you want speed and half the price of Opus.
- Bulk workHaiku 5.5 for classification, extraction and routing, the jobs you run ten thousand times.
Clodhopper move: Sending every request to the most expensive model because it “feels safer.” Capability you do not need is cost and waiting.
Thinking is adaptive on all four: the model decides how much to think, steered by an effort setting. On Fable 5.2 and Opus 5.5 it is always on.
Older models still available
Fable 5.1, Fable 5, Opus 5, Opus 4.8, Opus 4.7, Opus 4.6, Opus 4.5, Sonnet 5, Sonnet 4.6 and Haiku 4.5 are listed as legacy models. If your code names one of them, check the deprecations page for its retirement date.
Source: Anthropic's models overview and pricing page.