Anthropic expands context and session capacity for Claude Opus 5.5

Opus 5.5 is the first model in Anthropic’s Claude 5.5 family.[1] The company designed it for long-running agentic coding and knowledge work.[2] Anthropic says it leads its benchmarks in agentic coding, computer use and knowledge work.[1] The listed specifications include a one-million-token context window and a maximum output of 128,000 tokens.[2] Those capabilities are relevant to teams weighing whether a model can handle extended tasks and large working contexts; the listed specifications describe limits, not a guarantee that any particular task will fit or succeed.

Anthropic set Opus 5.5 API prices at $4 per million input tokens and $20 per million output tokens.[1] The company says the model costs 40% less to run than Opus 5.[1] Cache reads are priced at $0.20 per million tokens, which Anthropic says is 60% below Opus 5.[1] Opus 5.5 is available through the Claude API, Amazon Bedrock, Google Cloud, Microsoft Foundry and Claude Platform on AWS.[2] Anthropic is also increasing five-hour usage limits on Pro, Max, Team and seat-based Enterprise plans.[1] The evidence gives no revised limits for those plans.

Anthropic says context per request grew 2.6 times and Claude works 3.3 times longer on each prompt, with more than 40% more model calls per prompt.[3] It also reports 68% fewer interruptions and a more than 50% decrease in input that misses the cache.[3] Developers using API keys or cloud providers can set a one-hour cache lifetime for longer sessions.[3] In Claude Code, forked subagents start from their parent’s cache rather than paying for the same context again, according to Anthropic.[3] Teams evaluating costs should distinguish these reported changes from the published per-token API prices.

Anthropic says Opus 5.5 generates output more than 30% faster than Opus 5.[1] The company also reports that an early tester completed a 680,000-line code migration with the model in less than a day.[1] That is a company-reported example, not a general performance guarantee. External evaluators including Frontier Design and METR tested the model before release, according to Anthropic.[1] METR said acceleration from the model would be slightly higher than for Fable 5.1, but that it was unlikely to fully automate AI R&D.[4] Anthropic says Claude Sonnet 5.5 and Claude Haiku 5.5 are due to follow in the coming weeks.[1]