Anthropic Unveils Claude Opus 5.5
Credit: Anthropic
Anthropic has unveiled Claude Opus 5.5, the first model in the new Claude 5.5 family. The company describes it as a new flagship designed for agentic programming, computer use, and complex analytical tasks. At standard settings, Opus 5.5 costs nearly 40% less than Opus 5 while generating responses more than 30% faster.
A key area of development for Opus 5.5 has been handling complex, long-running tasks that require the model to independently plan its workflow, utilize tools, verify intermediate results, and make modifications to large-scale projects.
According to the developer, an early user leveraged Claude Opus 5.5 to migrate a codebase of 680,000 lines in under 24 hours—a task estimated to take an engineering team several weeks if performed manually. In another test, the model audited and corrected a 200,000-line codebase in less than three hours; the previous version, Claude Opus 5, required over 20 hours and 2.5 times as many tokens to complete the same task.
In a test involving the migration of HAProxy from C to Rust, the Opus 5.5 model completed the task in 9.5 hours—compared to 12 hours for Fable 5.1—at a 51% lower cost. Anthropic also highlights improvements in software optimization: in one trial, Opus 5.5 reduced page load times for the web application in 39 out of 40 instances, whereas Opus 5 tended to make less significant changes and sometimes altered the application’s behavior in the process.
According to Anthropic’s internal tests, Claude Opus 5.5 achieved a score of 66.4% on the Terminal-Bench 4.0 benchmark, which evaluates the ability of AI agents to perform multi-step tasks in the command line. This surpasses the results of Claude Opus 5, Fable 5.1, GPT-6 Astra, and GPT-5.6 Sol. On FrontierCode v1.1, the new model scored 54.4%, whereas Claude Fable 5.1 scored 50.3% and GPT-6 Astra scored 53.3%. In the CursorBench 4.0 benchmark, Opus 5.5 achieved a score of 57.8%, outperforming Fable 5.1 and Opus 5.
Claude Opus 5.5 is already available on all major platforms. In the coming weeks, the company promises to release Claude Sonnet 5.5 and Claude Haiku 5.5, featuring some of the same performance, efficiency, and safety improvements.
