XNews

Claude Opus 5.5 released: Fable-level performance at lower cost

Claude Opus 5.5 released: Fable-level performance at lower cost

Anthropic released Claude Opus 5.5, achieving Fable 5.1-level performance on most tasks while reducing cost by 40% for the same work. It is 30% faster, with input $4 and output $20 per million tokens. Cache read dropped from $0.5 to $0.2. The model also improved writing style and research report accuracy.

Key insights

Performance and Cost

Claude Opus 5.5 matches Fable 5.1 on most benchmarks at 40% lower cost. Input $4/M tokens, output $20/M tokens, 20% cheaper than Opus 5. Cache read drops 60% to $0.2/M tokens, benefiting heavy agent users. Speed increased by 30%.

Real-World Coding Efficiency

An early tester audited and fixed a 200k-line codebase in under 3 hours; Opus 5 required 20+ hours with 2.5x tokens. In internal tests, Opus 5.5 rewrote HAProxy from C to Rust in 9.5 hours, cost 51% less than Fable 5.1, passing nearly all regression tests.

Comparison with GPT-6 Astra

On Terminal-Bench 4.0, Opus 5.5 matches GPT-6 Astra in score but costs about 40% as much. Anthropic notes that at this level, benchmark differences no longer reflect real-world use.

Research Report Accuracy

In a test where models wrote quarterly earnings reports from web search, Opus 5.5 produced 16 of 18 hallucination-free reports, while Fable 5.1 and Opus 5 failed all 18.

Improved Writing Style

Opus 5.5 places the most important conclusion first and follows user writing instructions better, addressing feedback that Opus 5 was verbose and jargon-heavy.

Safety and Usage Limits

Due to capabilities near Mythos 5.1, Opus 5.5 inherits Fable 5.1's safety restrictions: most cybersecurity tasks are routed to older Opus 4.8; biology research is limited. Developers note that thinking mode cannot be turned off, and API accounts registered after Aug 31 cannot extract reasoning via history modification.

Availability and Future Releases

Opus 5.5 is available on Claude platform, AWS, Google Cloud, and Azure, API model name claude-opus-5-5. Pro, Max, Team users get increased five-hour usage limits and an extra reset. Sonnet 5.5 and Haiku 5.5 will follow in coming weeks.

Source posts