News · September 22, 2026
Claude Opus 5.5 lowers the cost of top-tier coding models
Anthropic released Claude Opus 5.5 with lower per-token prices than Opus 5 and higher scores on its coding and agent benchmarks. For teams running coding agents at scale, cost per task is the number to re-check.
What happened
- Anthropic announced Claude Opus 5.5 on September 22, 2026, available through its own platform and on Amazon Web Services, Google Cloud and Microsoft Azure.
- According to Anthropic, prices are $4 per million input tokens and $20 per million output tokens, down from $5 and $25 for Opus 5, and typical workloads at default settings cost about 40% less to run.
- Anthropic reported higher scores than Opus 5 on agentic coding and computer-use benchmarks, including Terminal-Bench 4.0 (66.4% vs 52.3%), and said the model produces output more than 30% faster.
- Anthropic also said the model resists prompt injection better than Opus 5 and that watermarking is applied to comply with the EU AI Act.
- TechNews reported that Anthropic puts Opus 5.5 close to its more capable Claude Fable 5.1 on most tasks, and that Sonnet 5.5 and Haiku 5.5 were expected within weeks.
Why it matters
Price cuts at the top tier change the routing math. Work that was moved to smaller models purely for cost, such as large refactors or migrations, may be worth running on the flagship again. Measure cost per completed task rather than cost per token, because tokens used per task change between model versions too.
Better resistance to prompt injection matters most for agents that read untrusted content such as tickets, emails and web pages. It helps, but it does not replace limiting what an agent can do with its tools and credentials.
With a new model generation every few months, keep model names in configuration rather than code, and run your own regression suite before switching. It is a routine step we build into AI projects from the start.
Sources
This post is CMM’s own summary and commentary on publicly reported news. Facts are attributed to the sources linked above; please read them for full details. CMM is not affiliated with these publishers or with the companies mentioned, and trademarks belong to their owners. Prepared with AI assistance. Corrections or requests: [email protected]. See our editorial policy.
Want help with this?
Tell us what you need, and we will suggest the right team model on a short call.
Talk to us