Anthropic Cuts Claude Opus 5.5 Prices 40%, Runs 30% Faster

Anthropic launched Claude Opus 5.5 today, a significant update to its flagship model that cuts costs by 40% on typical workloads while delivering faster output and stronger performance on the long-running, multi-step tasks that matter most to enterprise teams.

For businesses that have been watching AI model costs closely, this is the kind of release that changes the math on large-scale deployments.

What Changed

The pricing reduction is real and substantial. Input tokens drop from $5 to $4 per million. Output tokens fall from $25 to $20 per million. Cache reads are cut by 60%, from $0.50 to $0.20 per million tokens. On a workload that makes heavy use of caching, those savings compound quickly.

Speed improved too. Opus 5.5 generates output over 30% faster than Opus 5, which matters when you are running agents that need to complete multi-step tasks within a session or respond to users in near-real time.

On performance, Anthropic reports strong gains across the areas that enterprise teams care about most. The model scores 1846 Elo on GDPval-AA v2.1, a benchmark testing performance against real-world professional tasks. That compares to 1735 for Fable 5.1 and 1708 for Opus 5. In agentic coding tests, one early tester completed a 680,000-line code migration in under a day, work the team estimated would have taken weeks.

Another tester reduced a complex task from 38 prompts over four days down to 11 prompts over three hours, a compression of effort that points to better reasoning efficiency, not just raw speed.

Built for Sustained Agentic Work

Anthropic positions Opus 5.5 specifically for long-running jobs: codebase-wide migrations, technical audits, document review at scale, and complex professional workflows that require maintaining context across many steps. This is distinct from the “quick query, quick answer” use case that most people picture when they think about AI assistants.

A model optimized for sustained agentic work also addresses one of the consistent complaints about earlier Opus versions: verbosity. Anthropic says the model now puts the most important information first and writes with less padding, a practical improvement that reduces output tokens and, by extension, cost per task.

The safety profile also improved. Anthropic says Opus 5.5 posted the best scores of any model to date on its automated behavioral audit, with specific improvements in resisting prompt injection attacks and avoiding hard-to-reverse actions. For enterprise deployments where agents have real system access, those properties matter.

What This Means for Business

AI model pricing has been the quiet brake on enterprise adoption. Teams that want to build sophisticated agentic workflows often run the numbers and conclude that the economics do not justify the investment at current scale. A 40% cost reduction does not just trim a budget line. It resets whether certain use cases are viable at all.

For a business running automated analysis workflows, document processing pipelines, or AI-assisted development at any meaningful volume, this pricing shift opens up workloads that were previously marginal. The combination of lower cost per token, faster output, and better caching economics means the breakeven point on building agentic systems just moved in your favor.

Opus 5.5 is available now through AWS, Google Cloud, and Microsoft Azure. Anthropic also announced that Sonnet 5.5 and Haiku 5.5 are coming in the next few weeks, suggesting the full model lineup is being refreshed on the same cost and performance trajectory.

If your team has been evaluating whether to build AI agent systems or expand existing ones, the timing is worth noting. The models are getting meaningfully cheaper and more capable at the same time, and that combination does not happen every quarter.

Enterprise DNA helps businesses design and deploy AI agent systems that deliver measurable ROI. If you want to explore what agentic AI could look like for your operations, book a discovery call with Sam.