Claude Opus 5.5

Claude Opus 5.5 is Anthropic's flagship model, released September 22, 2026 as the first model in a new Claude 5.5 family. Anthropic says it performs at the level of Claude Fable 5.1 on most work while costing 40% less to run than Claude Opus 5. The API id is claude-opus-5-5. It keeps the 1M-token context window, 128K max output, and adaptive thinking, and it became the default model for Claude paid plans on day one.

Its main OpenAI rival is GPT-6 Sol, announced the same day.

Key Features

• Fable-level results at Opus prices: Input is $4 and output is $20 per 1M tokens, 20% below Opus 5. Cache reads drop 60% to $0.20 per 1M tokens, which matters most in long agentic and coding runs.
• 40% cheaper per typical workload: The saving comes from both the lower list price and fewer tokens consumed per task, and Anthropic says output is more than 30% faster than Opus 5.
• Adaptive thinking with a medium default: Thinking is always on and adaptive, and the default effort is medium. Anthropic recommends raising effort only when your own evals still fall short.
• Best score on Anthropic's behavioral audit: Anthropic reports Opus 5.5 as its strongest model yet on its automated behavioral audit, less likely to take hard-to-reverse actions, and more resistant to prompt injection than Opus 5.
• Day-one Copilot support: GitHub's supported-models table lists Claude Opus 5.5 as GA alongside Claude Fable 5.1 and Claude Opus 5.

Model Specifications

Pricing

From the Claude models overview, per 1M tokens:

Fast mode in Claude Code and the Claude Platform runs up to 2.5x faster at $8 per 1M input and $40 per 1M output. Anthropic also raised five-hour usage limits on Pro, Max, Team, and seat-based Enterprise plans and gave subscription users a rate-limit reset they can save and spend later.

Benchmarks

Anthropic's own comparison, with Opus 5.5 run at max effort unless noted:

AutomationBench at 40.0% was run by Zapier, and Anthropic notes safeguard interventions counted as failures there, so the number understates real results. Treat all of these as vendor or partner evaluations rather than independent audits.

Use Cases

• Codebase-wide migrations: Anthropic cites early testers doing a 680,000-line migration in under a day, and auditing and fixing a 200,000-line codebase in under three hours where Opus 5 took more than 20 hours and 2.5x the tokens.
• Long-running agents: The 1M context window and cheaper cache reads target multi-hour sessions in Claude Code.
• Knowledge work: Anthropic reports GDPval-AA v2.1 at 1846, ahead of Fable 5.1 and the GPT-6 models listed in its table.
• Regulated deployments: Opus 5.5 is deployed with Fable 5.1-level safeguards, with the Life Sciences Verification Program open today and the Cyber Verification Program expanding in the coming weeks.

Limitation: Anthropic itself warns that at this capability level, benchmark margins are a weaker guide to real-world differences, and that its gap to Fable 5.1 is narrower than the scores suggest. There is also a text watermark for EU AI Act compliance, and Anthropic's own factuality test still failed 2 of 18 tasks.

Getting Started

1. Set model to claude-opus-5-5 in the Claude API or pick it in Claude Code.
2. Stay at the default medium effort, then compare against xhigh and max on your own evals before paying for more thinking.
3. If you are migrating from Opus 5, expect lower cost per task rather than a hit to quality.
4. Use prompt caching. At $0.20 per 1M cache-read tokens, agentic sessions benefit the most.

Frequently Asked Questions

How is it different from Claude Fable 5.1?

Anthropic positions Opus 5.5 as level with Fable 5.1 on most work at much lower cost, and recommends Fable 5.1 only when your evals on Opus 5.5 at higher effort still fall short.

Is Claude Sonnet 5.5 out?

Not yet. Anthropic says Claude Sonnet 5.5 and Claude Haiku 5.5 will follow in the coming weeks with many of the same improvements.

Can I use it in GitHub Copilot?

Yes. GitHub lists Claude Opus 5.5 as GA in its supported AI models table, so it arrived in Copilot on the same day as its launch.

Alternatives

• Claude Fable 5.1: Anthropic's model for the most demanding reasoning.
• Claude Opus 5: the previous Opus flagship, still available at $5/$25.
• GPT-6 Sol: OpenAI's same-day release at $2/$10.

Tips

1. Budget on cost per task, not per token. The 40% saving Anthropic quotes combines a lower price with fewer tokens per task.
2. Measure your own evals before moving to Fable 5.1. Anthropic frames Opus 5.5 as the default starting point for most workloads.
3. Watch the safeguard paths. Tasks that trip the cyber or biology safeguards may complete on Opus 4.8 or Opus 5 instead, which changes both cost and answers.

Conclusion

Claude Opus 5.5 is Anthropic's new default frontier model: Fable-5.1-level capability on most work, a 1M context window, adaptive thinking, and $4/$20 pricing that undercuts Opus 5 by 20% on tokens and 40% on typical workloads. Read the announcement and the models overview, then benchmark it against GPT-6 Sol on your own tasks.