Anthropic has introduced Claude Opus 5.5, the first release in its new Claude 5.5 family, combining advanced coding and knowledge-work capabilities with a significant reduction in operating costs. The company says the model delivers performance broadly comparable to its more advanced Claude Fable 5.1 model on many types of work, while costing substantially less to operate.
A More Efficient High-End AI Model
Opus represents Anthropic’s highest-capability Claude tier, designed for demanding tasks that require extended reasoning, software development, research and complex workflows.
With Opus 5.5, Anthropic is focusing not only on intelligence but also on efficiency. The company reports that the model requires less computing power than Opus 5 and generates responses more than 30% faster.
The pricing has also been reduced. Input tokens cost $4 per million, while output tokens are priced at $20 per million. Cache reads cost $0.20 per million tokens. Anthropic says these changes result in approximately 40% lower costs on typical workloads compared with Opus 5.
For businesses using AI agents at scale, lower token consumption and reduced infrastructure costs could make complex AI workflows more economical.
Stronger Performance for Coding and Knowledge Work
Anthropic is positioning Opus 5.5 as a model particularly suited to large, multi-step assignments.
The company reports improvements across agentic coding, computer use and knowledge work benchmarks. In one example, an early tester used Opus 5.5 to handle a 680,000-line code migration in less than a day, a project Anthropic says would traditionally require weeks of engineering effort.
Another internal test involved improving the performance of a web application. Opus 5.5 successfully reduced page load times in 39 of 40 attempts, according to Anthropic.
The model is also designed for sprawling engineering projects, including codebase migrations, audits and software modernization.
Lower Costs Could Matter for AI Agents
One of the most important parts of the release is not simply benchmark performance. It is the relationship between capability and cost.
Anthropic says Opus 5.5 can complete demanding coding tasks while using fewer tokens than its predecessor. In an internal comparison involving HAProxy, the model completed a C-to-Rust migration in 9.5 hours compared with 12 hours for Fable 5.1, while costing 51% less in that test.
This could be particularly relevant for companies building autonomous coding agents, software development assistants and other AI systems that make hundreds or thousands of model calls.
A More Direct Communication Style
Anthropic has also changed how Opus 5.5 communicates.
The company says early testers found the new model clearer and easier to follow. Rather than burying the key answer beneath lengthy explanations, Opus 5.5 is designed to place important information earlier in its responses and follow user-provided writing instructions more closely.
For professional users, this could make long AI-assisted work sessions easier to review and manage.
Safety Remains a Major Part of the Release
The launch also comes during Anthropic’s broader effort to balance advances in AI capabilities with safety measures.
Opus 5.5 underwent pre-release evaluation involving external organizations including METR and Frontier Design. Anthropic says its automated behavioral audit showed the model performing better than previous models across thousands of simulated scenarios. The company also reports improvements in resistance to prompt injection and actions that could cross defined boundaries.
Because Anthropic considers Opus 5.5 comparable to Claude Mythos 5.1 in certain biology and cybersecurity capabilities, it is deploying safeguards similar to those used with its highest-capability systems.
Access to some sensitive capabilities will therefore be controlled through verification programs for areas such as life sciences and cybersecurity.
What Comes Next for Claude
Opus 5.5 is also notable because it follows Anthropic’s recent public discussion about slowing the pace of frontier AI development so that safety systems can keep up with increasingly capable models.
The company says future Claude releases will continue to receive improvements in performance, efficiency and safety.
Anthropic has also confirmed that Claude Sonnet 5.5 and Claude Haiku 5.5 are expected in the coming weeks, extending the new generation beyond its premium Opus tier.
The Bigger Picture
Claude Opus 5.5 represents a broader direction in the AI industry: better models are increasingly being judged not only by what they can accomplish, but also by how efficiently and safely they can do it.
For developers and enterprises, the combination of advanced coding capabilities, faster responses and lower operating costs could make sophisticated AI agents more practical for everyday workflows.
As competition among leading AI companies continues, efficiency may become just as important as raw benchmark performance.
