Claude Opus 5.5 Frontier Agentic Model
Released by Anthropic on 22 September 2026 as the first model in the Claude 5.5 family: Anthropic says it performs at the level of Claude Fable 5.1 on most work at 40% lower cost than Opus 5 on typical workloads
Tool Overview
Features, steps and FAQ below
Features
- ✓ Performance: Anthropic calls it a major step up from Opus 5 and its new leading model, with one early tester completing a 680,000-line code migration in less than a day
- ✓ Cost and speed: $4 input and $20 output per million tokens (20% less than Opus 5), $0.20 cache reads (60% less) and over 30% faster output, netting 40% lower cost on typical workloads at default settings
- ✓ Coding: 66.4% on Terminal-Bench 4.0 and 57.8% on CursorBench 4.0, with Anthropic saying it beats GPT-6 Astra at default effort on FrontierCode at roughly 20% of the cost per task
- ✓ Safety: the strongest model to date on Anthropic's automated behavioral audit, less likely to take hard-to-reverse actions and more resistant to prompt injection, with a full Opus 5.5 system card
- ✓ Deployment: biology and cybersecurity safeguards similar to Fable 5.1 with verification programs for vetted organizations, plus preserved-thinking anti-distillation protection
How to Use
- Set the model to claude-opus-5-5 on the Claude Platform
- Enable Fast mode in Claude Code and the Claude Platform for up to 2.5x speed when you need it
- Hand it long-horizon work such as codebase audits, migrations and web performance fixes, using cache reads to cut cost
- Use it on AWS, Google Cloud, Microsoft Azure or in the Claude app, and read the official system card to understand its limits
FAQ
What is Claude Opus 5.5?
A model released by Anthropic on 22 September 2026 and the first member of the Claude 5.5 family.
How does it perform?
Anthropic says it performs at the level of Claude Fable 5.1 on most work and is a major step up from Opus 5, with early testers seeing large gains on their most complex work.
How much does it cost?
Input and output tokens are $4 and $20 per million (20% less than Opus 5) and cache reads are $0.20 per million (60% less); Anthropic says typical workloads cost 40% less than Opus 5 at default settings.
What kinds of tasks does it suit?
Anthropic says it is particularly good at long, sprawling jobs like codebase-wide migrations and audits; an early tester audited and fixed a 200,000-line codebase in under three hours where Opus 5 took over 20 hours.
How safe is it?
Anthropic says it achieves the best scores of any model to date on its automated behavioral audit, is less likely to take hard-to-reverse actions and is more resistant to prompt injection, with details in the system card.
Where and when is it available?
It is available from launch on the Claude Platform and on AWS, Google Cloud and Microsoft Azure; Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks.