Strengths creators can use
- One context size across the range: 1M tokens in and 128K out on all three models means a long document set does not force you upward into a slower, costlier tier. You can keep the same material and change only the model, which is what makes comparing them on your own task practical.
- Effort is a dial before it is a purchase: Anthropic reports Fable 5.1 at low or medium effort matching or beating Fable 5 at much lower cost. Since thinking is adaptive and always on for Fable 5.1 and Opus 5.5, turning effort down is usually the first thing to try — before deciding you need a different model.
- Prices fell faster than capability: The 5.5 generation arrived with cuts rather than increases: Haiku 5.5 is priced 90% below Haiku 4.5 for prompts up to 100K tokens and 50% below above that; Opus 5.5 costs 40% less to run than Opus 5 with cache reads 60% cheaper; Fable 5.1 is an estimated 25% cheaper than Fable 5.
- Agentic coding leads the published rows: Opus 5.5 tops Anthropic's own agentic-coding table at 66.4% on Terminal-Bench 4.0, with Fable 5.1 at 55.8% and both ahead of the GPT-6 models Anthropic benchmarked. For long code work that is a real capability signal, not a marginal one.
- Charts and screenshots are readable: 89.0% on Chartography with tools for Opus 5.5 means a dashboard capture or a rendered chart can be treated as an input rather than a picture you have to transcribe first. That makes it usable for reporting and analysis work, not only for code.
- A small model that is actually fast: Haiku 5.5 is Anthropic's fastest model to date and is positioned as a subagent for coding alongside the larger Claude models. That combination — delegated subagent plus a real speed advantage — is what makes multi-step workflows affordable.
