Developers gain a model that finishes multi-hour engineering work in roughly one-seventh the time of its predecessor while spending far fewer tokens. Claude Opus 5.5, released by Anthropic, is now available across the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure under the model name claude-opus-5-5. The release pairs lower API pricing and faster output with watermarking designed to comply with the EU AI Act and new safeguards for sensitive work.
Built for Long and Complex Tasks
Opus 5.5 is designed for codebase migrations, software audits, financial analysis, data collection and workflows that span several applications. Early testers put it to work on engineering jobs that ran for hours at a time. In one documented test, Opus 5.5 audited and fixed a 200,000-line codebase in under three hours. Opus 5 needed more than 20 hours for the same task and used 2.5 times as many tokens.
Clio, a legal technology company, assigned the model a large engineering task across six repositories and let it run overnight unattended. The model stayed on task for over 18 hours, defining how the services communicate and working out how each one should apply the result. The account from Clio’s development team was that milestones arrived faster than with Opus 5 and required minimal reworking. The code comments were short and useful rather than long and prose-heavy.
Terminal-Bench 4.0, a benchmark that measures how well a model completes complex multi-step professional tasks inside a command line interface, was among the evaluation tools cited. Early testers also reported improvements in maintaining context, delegating work to other agents and checking results. Their accounts suggest the model can reduce the number of prompts, tool calls and corrections needed to finish a task.
What Changed in Coding and Knowledge Work
Anthropic reports gains in coding, computer use and professional knowledge work. Benchmarks and customer tests indicate that Opus 5.5 completes many tasks with fewer tokens than Opus 5, though performance varies according to the task, tools and effort settings.
Deloitte Consulting tested the model on code review and US consulting analysis. At its lowest effort setting, Opus 5.5 caught 72% of known bugs in code reviews, while Opus 5 at high effort caught 56%, with fewer false alarms and a fraction of the output. On consulting analysis, low thinking effort matched the higher thinking settings on half the output and passed Deloitte’s quality checks, which means more lower-effort deployments can be put into production and still deliver client-ready work.
Thinking Mode Stays On
Opus 5.5 cannot be used with thinking switched off. The model always runs a reasoning process, and developers can adjust how much effort it applies to a task. Anthropic says the model is easier to understand, places key information earlier in responses and follows writing instructions more closely.
The release also introduces preserved thinking, a safeguard that stops API users from editing the model’s prior context. Anthropic says the measure makes it harder to extract and copy the model’s capabilities through large-scale distillation attacks. Preserved thinking applies to Opus 5.5 API accounts created on or after August 31, 2026.
Additional Controls for Sensitive Work
Opus 5.5 includes safeguards covering cybersecurity, biology and attempts to copy the model. Most cybersecurity tasks are rerouted to Opus 4.8, and Anthropic plans to expand its Cyber Verification Program to give verified cybersecurity professionals broader access to Opus 5.5. Organizations whose biological research is impeded by the safeguards can apply to the Life Sciences Verification Program.
Anthropic added a classifier that screens coding-agent actions before execution. The company also introduced an open-source sandbox that security teams can audit, plus code-review features designed to catch vulnerabilities before changes are merged. The model has stronger defences against prompt-injection attacks. In an evaluation of attempts to cross containment boundaries, Opus 5.5 tried to circumvent its assigned limits about 85% less often than Opus 5 or Claude Mythos 5.1. Anthropic said every attempt was low severity and self-reported.
External organizations, including METR and Frontier Design, evaluated the model before release. Anthropic acknowledges that current evaluations cannot identify every potential failure before deployment.
Lower API Prices and Faster Output
Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Opus 5. Cached input reads cost $0.20 per million tokens, down from $0.50. The lower cache price benefits coding agents and other systems that consult the same instructions, files or conversation history repeatedly.
The model generates output more than 30% faster than Opus 5. A separate fast mode is available through Claude Code and the Claude Platform, offering up to 2.5 times the speed for $8 per million input tokens and $40 per million output tokens. Anthropic is increasing five-hour usage limits for Pro, Max, Team and seat-based Enterprise customers. Subscription users will receive a rate-limit reset that they can save and use later.
FAQ
How much does Claude Opus 5.5 cost?
Claude Opus 5.5 costs $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Opus 5. Cached input reads cost $0.20 per million tokens, down from $0.50. A fast mode costs $8 per million input tokens and $40 per million output tokens.
How much faster is Claude Opus 5.5 than Opus 5?
Opus 5.5 generates output more than 30% faster than Opus 5. Its fast mode offers up to 2.5 times the speed. In one test, Opus 5.5 audited and fixed a 200,000-line codebase in under three hours, while Opus 5 took more than 20 hours and used 2.5 times as many tokens.
Can developers turn off thinking mode in Claude Opus 5.5?
No. Opus 5.5 cannot be used with thinking switched off. The model always uses a reasoning process, and developers can adjust how much effort it applies to a task.
This article summarizes reporting from helpnetsecurity.com.