Anthropic releases Claude Opus 5.5 as the first model in its new Claude 5.5 family. The company says the model delivers performance comparable to Claude Fable 5.1 on most work while costing 40% less to run than Opus 5. The new model focuses on agentic coding, computer use, knowledge work, speed, and safety. Anthropic also gives Opus 5.5 stronger safeguards and tests it with external safety organizations before release.
Claude Opus 5.5 Targets Complex AI Work
Claude Opus 5.5 is designed for tasks that require multiple steps, long sessions, and the use of tools. Anthropic says the model leads its benchmark results in agentic coding, computer use, and knowledge work. The company reports that Opus 5.5 performs especially well on large software projects. In one early test, a user gives the model a 680,000-line code migration, and it completes the work in less than a day.
Another test involves a 200,000-line codebase, which Opus 5.5 audits and fixes in under three hours. Anthropic also reports that Opus 5.5 succeeds 39 out of 40 times when asked to reduce page load times across a web application. These results show the model’s focus on completing larger engineering tasks instead of handling only small coding requests.
Better Coding Performance and Efficiency
Coding is one of the main areas where Anthropic positions Claude Opus 5.5. The model works on code migrations, audits, debugging, code reviews, and other multi-file development tasks. On Terminal-Bench 4.0, Opus 5.5 scores 66.4%, compared with 52.3% for Opus 5. On FrontierCode, it scores 54.4%, while Opus 5 scores 48.0%. It also reaches 57.8% on CursorBench, compared with 46.6% for Opus 5.
The model also uses fewer tokens and requires fewer steps for many tasks. Anthropic says an internal HAProxy migration from C to Rust takes Opus 5.5 about 9.5 hours, compared with 12 hours for Fable 5.1, while costing 51% less. This efficiency can make a difference for developers who run AI agents for long periods. Lower token usage means teams can run more tasks without increasing their AI spending at the same rate.
Lower Pricing and Faster Responses
Anthropic also makes cost a major part of the Claude Opus 5.5 release. The model costs $4 per million input tokens and $20 per million output tokens. Opus 5 costs $5 per million input tokens and $25 per million output tokens. Cache reads cost $0.20 per million tokens, compared with $0.50 for Opus 5. Anthropic says that its typical workloads cost 40% less to run because Opus 5.5 both uses less compute and completes tasks with fewer tokens.
The model also generates output more than 30% faster than Opus 5. A fast mode is available in Claude Code and the Claude Platform, with speeds of up to 2.5 times the standard rate. This option comes with higher token pricing. For businesses using AI agents, the combination of lower costs and faster responses can make longer automated workflows easier to operate.
Stronger Safety and Prompt Injection Protection
Anthropic gives safety a major role in the Opus 5.5 release. The company says the model achieves its strongest result so far on its automated behavioral audit, which tests Claude across thousands of simulated scenarios. Opus 5.5 also receives stronger protection against prompt injection attacks. Anthropic says it matches or beats Opus 5 across the coding, tool-use, computer-use, and web-browsing settings it tests. A benchmark from AI security firm Gray Swan also shows Opus 5.5 tying Fable 5.1 for the lowest prompt injection success rate among the models tested.
Before launch, external evaluators including Frontier Design and METR test the model. Anthropic also says Opus 5.5 is about 85% less likely than Opus 5 or Mythos 5.1 to attempt to bypass containment boundaries in a dedicated internal evaluation. Because of its capabilities in areas such as biology and cybersecurity, Anthropic applies safeguards similar to those used for Fable 5.1. Vetted researchers can also apply for access through Anthropic’s verification programs.
Claude Opus 5.5 Improves Knowledge Work
The new model is not limited to software development. Anthropic also focuses on research, financial analysis, business workflows, and other professional tasks. On GDPval-AA v2.1, which evaluates professional work across 44 occupations, Opus 5.5 scores 1,846 Elo at maximum effort. Fable 5.1 scores 1,735, while Opus 5 scores 1,708. Anthropic also reports strong results on business workflow and large-scale data collection evaluations.
The company tests Opus 5.5 on a financial analysis task that requires creating an Excel model and an executive presentation for a fictional merger. The model completes the task in 63 minutes, compared with 93 minutes for Opus 5, and costs 50% less to produce. Opus 5.5 also improves how it communicates. Anthropic says the model puts important information earlier, uses less jargon, and follows writing instructions more closely. This makes its responses easier to review during long work sessions.
Claude 5.5 Family Expands
Claude Opus 5.5 is the first release in Anthropic’s new Claude 5.5 family. The company says Claude Sonnet 5.5 and Claude Haiku 5.5 follow in the coming weeks with similar improvements in performance, efficiency, and safety.
Opus 5.5 is available through the Claude Platform and major cloud providers, including Amazon Web Services, Google Cloud, and Microsoft Azure. Claude AI with MCP also supports connections between Claude and external AI tools. Anthropic’s latest release combines stronger performance with lower operating costs. Its focus on coding, knowledge work, agentic tasks, and safety makes Claude Opus 5.5 another significant step in the company’s model lineup.
