Claude Opus 5.5 Release Notes
Claude Opus 5.5 delivers higher performance at 40% lower cost
Anthropic has introduced Claude Opus 5.5, the first model in the Claude 5.5 family. The model is designed to match the performance of Claude Fable 5.1 on most tasks while reducing the cost of running typical workloads by 40% compared to its predecessor, Opus 5. It also generates output more than 30% faster than Opus 5.
Pricing and Efficiency
Opus 5.5 is more compute-efficient than Opus 5, resulting in lower token pricing and reduced token usage per task.
| Prices per 1M tokens | 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 |
For users requiring higher speeds, a "Fast mode" is available in Claude Code and the Claude Platform, offering up to 2.5x speed at $8 per million input tokens and $40 per million output tokens.
Advancements in Agentic Coding and Software Engineering
Claude Opus 5.5 shows significant improvements in handling large-scale, complex software engineering tasks, specifically in codebase-wide migrations and audits.
Real-world Coding Performance
- Large-scale migrations: One early tester completed a 680,000-line code migration in less than a day.
- Codebase audits: In one test, Opus 5.5 audited and fixed a 200,000-line codebase in under three hours, whereas Opus 5 required over 20 hours and 2.5x more tokens.
- Language translation: In a test translating HAProxy from C to Rust, Opus 5.5 finished in 9.5 hours (compared to 12 hours for Fable 5.1) and cost 51% less.
Benchmarks
Opus 5.5 leads in agentic coding and computer use. On Terminal-Bench 4.0, it matches GPT-6 Astra at approximately 40% of the cost. On CursorBench, it outperforms GPT-5.6 Sol by 11 points at roughly one-third of the cost.
Knowledge Work and Research Capabilities
Opus 5.5 is optimized for high-reliability research and financial analysis, reducing hallucinations in data-heavy reports.
Research Accuracy
In an internal test requiring a report based on hard-to-locate earnings releases, 16 out of 18 Opus 5.5 reports passed a quality bar that required zero invented figures or quotes. Neither Fable 5.1 nor Opus 5 cleared this bar in any attempt.
Business and Financial Analysis
- Financial Modeling: In a fictional merger analysis, Opus 5.5 produced a more thorough financial model and a more readable executive presentation than Opus 5, finishing in 63 minutes (vs 93 minutes) and costing 50% less.
- Professional Work: On the GDPval-AA v2.1 benchmark (covering 44 occupations), Opus 5.5 scored 1846 Elo, surpassing both Fable 5.1 (1735) and Opus 5 (1708).
Communication and Collaboration Style
Anthropic has overhauled the communication style of Opus 5.5 to be more natural and less verbose. The model now prioritizes the most important information at the beginning of its responses and avoids idiosyncratic jargon, making it a more effective partner for long working sessions.
Safety, Alignment, and Safeguards
Opus 5.5 is deployed with a focus on "pacing the frontier," ensuring safety practices evolve alongside model capabilities.
Alignment and Behavioral Audit
Opus 5.5 achieved the best scores to date on Anthropic's automated behavioral audit. It is significantly more resistant to prompt injection and less likely to take hard-to-reverse actions or circumvent containment boundaries. In tests of containment boundary crossing, it attempted to circumvent boundaries 85% less often than Opus 5 or Claude Mythos 5.1.
Domain-Specific Safeguards
Due to its high capabilities in biology and cybersecurity, Opus 5.5 employs safeguards similar to those of Fable 5.1:
- Cybersecurity: Most cybersecurity tasks are re-routed to Opus 4.8. Verified practitioners can apply for the Cyber Verification Program for expanded access.
- Biology: Biology-related work is managed via the Life Sciences Verification Program for vetted organizations.
- Distillation: To prevent the extraction of model capabilities, Opus 5.5 uses "preserved thinking," which prevents API users from editing prior context to extract reasoning.
Availability
Claude Opus 5.5 is available on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can access it via the Claude Platform using the model identifier claude-opus-5-5.