Anthropic has released the Claude Opus 5.5 model and reduced API pricing by 20% compared to Opus 5. OpenAI introduced GPT-6 Sol and GPT-6 Luna, lowering base API rates for Sol by 50%.

Opus 5.5 Outperforms GPT-5.6 Sol and GPT-6 Astra in Benchmarks

The Claude Opus 5.5 release highlights improvements in complex coding, handling large projects, and multi-step tasks. According to published benchmarks, the model outperforms GPT-5.6 Sol in all 6 general tests and GPT-6 Astra in 4 out of 6.

GPT-6 Sol for Complex Coding and Agents; Luna for Broad Use Cases

OpenAI introduced GPT-6 Sol and Luna: Sol is designed for complex coding and agent tasks, while Luna targets simpler, high-volume scenarios. In the AutomationBench work tasks test, Sol outperforms Claude Opus 5 and Fable 5.1, with task execution estimated to be about 11 times less expensive than Opus 5. According to the company’s internal measurements, actual errors have been reduced by half compared to GPT-5.6 Sol.