Anthropic Cuts Claude Opus 5.5 Costs by 40 Percent

Anthropic has released Claude Opus 5.5, a model that delivers a rare combination in the artificial intelligence market: increased capability alongside reduced pricing. The new version outperforms the previous Opus 5 in key areas while costing significantly less to operate, shifting the value proposition for developers and heavy users.

This release marks a strategic departure from the typical trend where higher performance comes with a steeper price tag. By lowering the cost per token, Anthropic is making advanced AI assistance more accessible for teams that process large volumes of data, effectively increasing their budget efficiency without sacrificing output quality.

Performance gains in coding and knowledge

According to data reported by XDA Developers, the new model shows measurable improvements in specific professional tasks. It achieves an eleven-percentage-point increase in agentic coding capabilities compared to the previous version, as measured by Terminal-Bench 4.0. Additionally, it records a six-percentage-point gain in task handling on CursorBench 4.0, indicating better precision in practical programming scenarios.

The model also scores higher in knowledge work evaluations, suggesting it can handle complex information retrieval and synthesis more effectively. Anthropic claims that for most general tasks, Claude Opus 5.5 now performs at a level similar to their top-tier Claude Fable 5.1. This suggests that the mid-range model has closed the gap with the premium option, offering near-flagship performance without the premium label.

Significant reduction in usage costs

The most tangible benefit for users is the price drop. Anthropic has reduced the input cost from five dollars to four dollars per million tokens. The output cost has similarly decreased from twenty-five dollars to twenty dollars per million tokens. This represents a forty percent reduction in expenses for organizations that rely heavily on generative AI for continuous operations.

Trade-offs for enterprise adoption

While the cost-performance ratio improves, the trade-off lies in the distinction between general tasks and highly specialized ones. Although Opus 5.5 matches the top-tier model for most workloads, it may still lack the extreme edge cases that the flagship version handles. For businesses, the decision now hinges on whether the forty percent savings justify the slight potential difference in ultra-complex reasoning tasks.

This move pressures competitors to reconsider their pricing structures, as the market increasingly demands high efficiency without proportional cost increases. For developers, the immediate impact is a higher ceiling for what their existing budgets can achieve, allowing for more iterative testing and larger-scale deployments. The release underscores a growing expectation that AI advancements should benefit both capability and affordability simultaneously.