Claude Opus 5.5 Released with Improved Performance and Cost

Anthropic released Claude Opus 5.5, matching Fable 5.1 performance on most tasks while costing 40% less than Opus 5. Input and output prices are $4 and $20 per million tokens, with cache reads reduced to $0.2. Output speed is 30% faster. Safety restrictions from Fable 5.1 are carried over.
Key insights
Claude Opus 5.5 Matches Fable 5.1 Performance at Lower Cost
Anthropic released Claude Opus 5.5, which matches the performance of its own Mythos-level Fable 5.1 on most tasks while costing 40% less than the previous Opus 5. Input and output token prices are $4 and $20 per million tokens, a 20% reduction. Cache reads dropped from $0.5 to $0.2, benefiting heavy agent users. Output speed increased by over 30%.
Real-World Examples Show Significant Efficiency Gains
An early tester used Opus 5.5 to audit and fix a 200,000-line codebase in under 3 hours, compared to 20+ hours with Opus 5 (2.5x more tokens). Internally, Opus 5.5 rewrote HAProxy from C to Rust in 9.5 hours at 51% lower cost than Fable 5.1's 12 hours. On Terminal-Bench 4.0, Opus 5.5 scored equal to GPT-6 Astra at about 40% of the cost.
Improved Factual Accuracy and Writing Style
In a test where models had to write quarterly earnings reports using only web search, Opus 5.5 produced 16 out of 18 reports without any fabricated numbers or quotes, while Fable 5.1 and Opus 5 failed all attempts. The model also adopts a more direct writing style, putting conclusions first and following user instructions more closely, addressing complaints about Opus 5's verbosity and jargon.
Safety Restrictions from Fable 5.1 Apply to Opus 5.5
Because Opus 5.5's capabilities in cybersecurity and biology approach Mythos 5.1, it inherits Fable 5.1's safety restrictions: most cybersecurity tasks are automatically routed to older Opus 4.8, and biology-related work is limited. Academic labs and pharma companies can apply for the Life Sciences Verification Program to unlock restrictions. A verification path for security practitioners will open in coming weeks.
API Changes: Thinking Mode Mandatory, Distillation Prevention
Developers should note two changes: thinking mode can no longer be disabled, and API accounts registered after August 31 cannot extract model reasoning via historical context manipulation. This is to prevent distillation (using many accounts to extract model capabilities). Existing integrations should check compatibility. Opus 5.5 is available on Claude platforms and AWS, Google Cloud, Azure as claude-opus-5-5.