On July 1, 2026, Anthropic made an official announcement regarding the launch of its latest large language model, Claude Sonnet 5. This new model is strategically positioned as a cost-effective option within the mid-to-high-tier range. Notably, it has substantially reduced the performance disparity with the flagship Opus series, emerging as the most 'agentic' iteration of the Sonnet model to date. Sonnet 5 showcases remarkable autonomy in task planning, enabling it to seamlessly call upon external tools like browsers and terminals, thereby achieving a high level of operational automation.

In the realm of programming, Sonnet 5 has demonstrated its prowess by scoring 63.2% on the SWE-benchPro, outperforming both GPT-5.5 and Gemini 3.5 Flash. This score places it just behind Opus 4.8, which achieved a 69.2% rating, securing Sonnet 5's position as the second-best performer in this category. Furthermore, in interdisciplinary reasoning benchmark tests, Sonnet 5 attained a score of 57.4%, narrowly trailing Opus 4.8 by a mere 0.5 percentage points.

When it comes to pricing, Sonnet 5 offers a promotional API pricing scheme that is valid until August 31. During this period, the cost is set at $2 per million input tokens and $10 per million output tokens. This pricing structure renders Sonnet 5 approximately 60% more economical than Opus 4.8. Even after the promotional period concludes, the price of Sonnet 5 will remain about 40% lower than that of Opus 4.8.