Claude Sonnet 5: strong agentic performance at a higher cost per task (artificialanalysis.ai)

Anthropic has announced the release of Claude Sonnet 5, which achieves a score of 53 on the Artificial Analysis Intelligence Index, matching the performance of GPT-5.5. While it shows notable improvements over its predecessor, Sonnet 4.6, in agentic task execution—utilizing approximately 40% more output tokens and performing significantly better in knowledge work evaluations—its pricing comes in higher than Opus 4.8. At $2.29 per task, which is a 2x increase from Sonnet 4.6, Claude Sonnet 5 is set to challenge cost-sensitive users within the AI/ML community, although promotional pricing may temporarily mitigate this. Despite the enhancements, Claude Sonnet 5 still lags behind in heavy reasoning tasks when compared to larger models like Opus 4.8 and GPT-5.5. Improvements have been notable across various benchmarks—it outperforms its predecessor and competes well with contemporaries in producing professional outputs—but the trade-off in cost and efficiency may steer some developers towards alternative models. With a context window of 1 million tokens and a refined effort setting configuration that includes an 'xhigh' rating, Claude Sonnet 5 represents a significant step forward while also raising questions about the balance between cost and capability in the evolving AI landscape.