XNews

Claude Opus 5.5 Released with Improved Performance and Cost Efficiency

Claude Opus 5.5 Released with Improved Performance and Cost Efficiency

Anthropic released Claude Opus 5.5, matching the performance of its flagship Fable 5.1 while reducing costs by 40%. The model is 30% faster and available across Claude, AWS, Google Cloud, and Azure. API pricing is set at $4 per million input tokens and $20 per million output tokens.

Key insights

Performance and Cost Efficiency

Claude Opus 5.5 matches the performance of Anthropic's flagship Fable 5.1 on most work tasks while costing 40% less than its predecessor Opus 5. Cost savings come from both lower per-token pricing and reduced token consumption for the same tasks. Output speed is also over 30% faster.

Real-World Coding Performance

In a test, Opus 5.5 audited and fixed a 200,000-line codebase in under 3 hours, a task that took Opus 5 over 20 hours with 2.5x the token usage. It also rewrote the HAProxy load balancer from C to Rust in 9.5 hours at 51% lower cost than Fable 5.1, passing nearly all regression tests.

Benchmarking and Research Capabilities

On the Terminal-Bench 4.0 benchmark for multi-step command-line tasks, Opus 5.5 scored equal to GPT-6 Astra at roughly 40% of the cost. In a research test requiring the model to write a quarterly earnings report using only web search, 16 out of 18 reports from Opus 5.5 passed without any fabricated numbers or quotes, while Fable 5.1 and Opus 5 failed all attempts.

Safety Restrictions and Alignment

Due to its capabilities approaching Mythos 5.1 in cybersecurity and biology, Opus 5.5 inherits Fable 5.1's safety guardrails. Most cybersecurity tasks are automatically routed to the older Opus 4.8, and biological research is restricted. A Life Sciences Verification Program is available for institutions to request elevated access. Thinking mode cannot be disabled.

LEGO Design Demonstration

A demonstration by the HuggingFace team showed Opus 5.5 designing a Microduck robot using 1,113 real LEGO parts in 6 colors. The model generated a 141-page, 237-step building manual with zero part collisions, along with LDraw model files, a CSV parts list, and a BrickLink XML wishlist, showcasing advanced spatial reasoning.

Availability and Future Releases

Opus 5.5 is now available on all Claude platforms (Pro, Max, Team) and through API on AWS, Google Cloud, and Azure. The API model is named claude-opus-5-5. Anthropic also announced that Sonnet 5.5 and Haiku 5.5 will be released in the following weeks.

Improved Writing Style

Anthropic stated that Opus 5.5 features a significantly improved writing style compared to Opus 5, which was often criticized for being verbose and using jargon. The new model places the most important conclusions first and adheres more closely to user-provided writing instructions.

Developer Benefits and Pricing Details

For developers, the cache read price dropped significantly from $0.50 to $0.20 per million tokens, a key cost for agentic workloads. Usage limits for Pro, Max, and Team subscribers were increased, and users received an extra limit reset. API accounts registered after August 31 cannot extract model reasoning via historical context to prevent distillation.

Source posts