Summary
Anthropic has introduced Claude Opus 5.5, the first model in its Claude 5.5 family. The company says it improves coding, knowledge work and safety performance while costing 40% less than Opus 5 on typical workloads.
Anthropic has introduced Claude Opus 5.5, the first model in its Claude 5.5 family. The company says the model performs at roughly the level of Claude Fable 5.1 on most work, while costing 40% less to run than Opus 5 on typical workloads.
Opus 5.5 is aimed particularly at long, multi-step tasks such as software migrations, code audits, business analysis and research. Anthropic also says it is faster, easier to follow in extended conversations and more resistant to prompt-injection attacks than its previous Opus model.
Lower-cost performance for coding and knowledge work
Anthropic reports that Opus 5.5 leads its evaluated models on agentic coding, computer use and knowledge-work benchmarks. “Agentic” tasks involve a model using tools and taking multiple steps to complete a larger objective rather than producing a single response.
On the company’s reported evaluations, Opus 5.5 scored 66.4% on Terminal-Bench 4.0, which measures complex command-line tasks, and 57.8% on CursorBench 4.0, which uses ambiguous, multi-file coding tasks based on real Cursor sessions. On GDPval-AA v2.1, a knowledge-work evaluation covering 44 occupations, it scored 1,846 Elo at maximum effort, compared with 1,735 for Fable 5.1 and 1,708 for Opus 5.
Anthropic says the model’s advantage is especially pronounced in efficiency. At default effort, it reported that Opus 5.5 beat GPT-6 Astra at maximum effort on some coding and knowledge-work comparisons at about one-fifth of the cost per task. On Terminal-Bench 4.0, the company says it matched GPT-6 Astra at about 40% of the cost, while on CursorBench it scored 11 points higher than GPT-5.6 Sol at about one-third of the cost per task.
The company’s examples focus on work that normally requires repeated tool calls. An early tester used Opus 5.5 on a 680,000-line code migration and completed it in less than a day, according to Anthropic. In another reported test, the model audited and fixed a 200,000-line codebase in under three hours, compared with more than 20 hours for Opus 5.
Anthropic says the model also uses fewer tokens per task. Its internal conversion of HAProxy from C to Rust passed nearly all of HAProxy’s regression tests with both Opus 5.5 and Fable 5.1, but Opus 5.5 completed the work in 9.5 hours rather than 12 and cost 51% less.
The company cautions that benchmark margins are becoming a less reliable guide to real-world differences between highly capable models. It says the practical gap between Opus 5.5 and Fable 5.1 was narrower in its own use than some benchmark scores indicated.
Pricing, safeguards and availability
Anthropic lists the following standard API prices per million tokens:
| Charge | Claude Opus 5.5 | Claude Opus 5 |
|---|---|---|
| Cache reads | $0.20 | $0.50 |
| Input tokens | $4 | $5 |
| Output tokens | $20 | $25 |
| Cache writes | $5 | $6.25 |
Cache reads are stored prompt or context material reused across requests. Anthropic says they make up a majority of the cost of many agentic and coding workloads. Opus 5.5 generates output more than 30% faster than Opus 5, according to the company.
A faster mode is available in Claude Code and the Claude Platform, with speeds of up to 2.5 times the standard mode. It costs $8 per million input tokens and $40 per million output tokens. Anthropic is also increasing five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans, and is adding a rate-limit reset that subscription users can save for later use.
Safety testing covered thousands of simulated scenarios in Anthropic’s automated behavioural audit. The company says Opus 5.5 was its strongest-performing model on that audit, was less likely than recent models to take hard-to-reverse actions or operate outside assigned boundaries, and was more resistant to prompt injection than Opus 5.
For coding deployments, Anthropic says Opus 5.5 includes a classifier that screens actions before execution, an open-source sandbox that security teams can audit and code review intended to catch vulnerabilities before changes are merged. Because Anthropic rates the model as comparable to Claude Mythos 5.1 in biology and cybersecurity, it is applying safeguards similar to those used for Claude Fable 5.1.
Vetted organisations can apply to use Opus 5.5 through Anthropic’s Life Sciences Verification Program. The company says access to its Cyber Verification Program will expand in the coming weeks for verified cybersecurity practitioners. Claude Sonnet 5.5 and Claude Haiku 5.5 are also planned for release in the coming weeks.
The reported performance results combine Anthropic’s own evaluations, early-access testing and external assessments. Results vary with model effort, tools, safeguards and task design; Anthropic notes that some benchmark tasks were completed by older models when Opus 5.5’s production safeguards intervened.