Anthropic unveils Claude Opus 5.5
- Maxime Hiez
- Anthropic
- 24 Sep, 2026
Introduction
Anthropic announces Claude Opus 5.5, the first model in the new Claude 5.5 family. It performs at the level of Claude Fable 5.1 on most tasks, at a cost 40% lower than Opus 5.
Check the Claude Opus 5 article HERE.
A model born from a deliberate slowdown in pace
Opus 5.5 is Anthropic’s first release since its CEO Dario Amodei called for pacing the frontier of AI models so that safety practices stay ahead of capabilities. The model was tested before release by external evaluators, including Frontier Design and METR, and scores the best result Anthropic has ever recorded on its automated behavioral audit, the company’s most comprehensive alignment test.
Performance
On reference evaluations, at maximum reasoning effort :
| Opus 5.5 | Fable 5.1 | Opus 5 | GPT-6 Astra | GPT-5.6 Sol | |
|---|---|---|---|---|---|
| Terminal-Bench 4.0 | 66.4% | 55.8% | 52.3% | 57.9% | 37.3% |
| GDPval-AA v2.1 | 1846 | 1735 | 1708 | 1542 | 1588 |
| AutomationBench | 40.0% | 31.4% | 26.9% | 41.4% | 28.8% |
| OSWorld 2.0 (partial) | 81.8% | 80.7% | 74.0% | n/a | n/a |
| Humanity’s Last Exam | 67.7% | 65.6% | 63.6% | 57.2% | n/a |
Anthropic tempers these gaps, however : at this level of capability, benchmark margins increasingly fail to reflect real-world differences. In its own internal use, the gap between Opus 5.5 and Fable 5.1 is narrower than these scores suggest.
Cost and speed
The price per million tokens drops across the board :
| Opus 5.5 | Opus 5 | |
|---|---|---|
| Input | 4$ | 5$ |
| Output | 20$ | 25$ |
| Cache (read) | 0.20$ | 0.50$ |
| Cache (write) | 5$ | 6.25$ |
A Fast mode, up to 2.5 times faster, is available in Claude Code and the Claude Platform at 8$ / 1M input tokens and 40$ / 1M output tokens. Opus 5.5 also generates responses more than 30% faster than Opus 5. Alongside this price cut, Anthropic is raising the five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans, and introducing a rate limit reset that users can save and trigger whenever they choose.
Coding
Opus 5.5 excels at long, sprawling jobs, codebase-wide migrations or audits. One tester audited and fixed a 200,000-line codebase in under three hours, compared with more than 20 hours and 2.5 times as many tokens for Opus 5. In an internal test translating HAProxy from C to Rust, Opus 5.5 finished in 9.5 hours compared with 12 for Fable 5.1, at a 51% lower cost.
- GitHub : Reports that Opus 5.5 “used among the fewest tokens and steps we measured” across GitHub Copilot CLI and VS Code, solving more terminal tasks than Opus 5 in less than half the steps.
- Stripe : Describes an Opus 5.5 session that directed a dozen sub-sessions on a 40-pull-request rebase spread over several days, with all 40 passing CI the next afternoon.
Knowledge work
On GDPval-AA v2.1, a test of real-world professional work across 44 occupations, Opus 5.5 scores 1846 Elo, ahead of Fable 5.1 (1735) and Opus 5 (1708). In an internal test writing a financial report limited to sources available on a copy of the web, 16 of Opus 5.5’s 18 reports cleared the quality bar set, against none for Fable 5.1 or Opus 5.
- Hebbia : Reports 86.6% coverage of expected criteria on end-to-end finance workflows, compared with 60.3% for Opus 5.
- Walleye Capital : States that Opus 5.5 detected an indexing error in its own evaluation instructions “and corrected for it, noting that this would cost it points with the grader”, an error no previous model had caught.
Communication
Anthropic reworked the way Opus 5.5 writes and communicates, one of the most frequent pieces of feedback on Opus 5. The model puts the most important information up front, limits jargon, and follows style instructions more faithfully.
- Ramp : Calls Opus 5.5 a “colleague” whose writing needs very little editing, even preferring its rewrite of a prompt over the original.
- Box : Reports that Opus 5.5 used a third of Opus 5’s tokens for responses 40% less verbose, with no loss of accuracy.
Alignment and safeguards
On its automated behavioral audit, which covers nearly 2,000 simulated scenarios, Opus 5.5 scores better than any recent Claude model on nearly every measure of misaligned behavior. In a new evaluation testing the propensity to cross containment boundaries, Opus 5.5 attempted to circumvent them about 85% less often than Opus 5 or Claude Mythos 5.1, with every attempt remaining low severity and self-reported.
Given its biology and cybersecurity capabilities are now comparable to Mythos 5.1’s, Opus 5.5 is deployed with safeguards similar to Fable 5.1’s :
- Cybersecurity : Most cybersecurity tasks are redirected to Opus 4.8. The Cyber Verification Program is expanding to Opus 5.5 in the coming weeks, with three tiers of trusted access.
- Biology : Eligible organizations, academic labs, startups, and pharmaceutical companies, can apply now to the Life Sciences Verification Program for dedicated research access.
- Anti-distillation : The preserved thinking mechanism, introduced with Fable 5.1, applies to Opus 5.5 for API accounts created on or after August 31, 2026.
info
Availability
Claude Opus 5.5 is available now on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. On the Claude Platform, developers can access it under the identifier claude-opus-5-5. The model remains available with zero data retention and includes the digital watermark required by the European AI Act. Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks, with much of the same improvements to performance, efficiency, and safety.
Conclusion
Opus 5.5 changes the Claude lineup’s hierarchy less than it redistributes its price-to-performance ratio : a level of performance close to Fable 5.1 at 40% less than Opus 5’s rate, with the best alignment score Anthropic has ever measured. For teams already in production on Opus 5, it’s a direct migration that lowers the bill without an apparent compromise on safety.
Sources
Did you enjoy this post ? If you have any questions, comments or suggestions, please feel free to send me a message from the contact form.
Don’t forget to follow us and share this post.
