OpenAI Launches GPT-6 Sol and Luna with 50% Cost Reduction

OpenAI has launched GPT-6 Sol and GPT-6 Luna, both models designed for professional work and coding tasks, with a significant 50% reduction in API pricing compared to GPT-5.6. The cost per task for GPT-6 Sol is set at $0.27, while GPT-6 Luna offers a 93% reduction in cost per task compared to Claude Opus 5 and 96% less than Claude Fable 5.1.

GPT-6 Sol achieves a maximum effort score of 68.8%, outperforming Claude Opus 5, which scores 26.9% at a cost that is 11.1 times higher. Notably, GPT-6 Sol makes half as many mistakes as its predecessor, GPT-5.6, and is approaching the reliability of GPT-6 Astra at a much lower cost. On the AutomationBench platform, GPT-6 Sol at xhigh effort outperforms Claude Opus 5 at max effort for just 9% of the cost per task.

GPT-6 Luna shows a 5.4 percentage point improvement in performance while maintaining a 58% lower cost per task compared to its predecessor. OpenAI has also enhanced prompt caching for GPT-6, resulting in over a 50% reduction in fresh processing requirements, which could lead to more efficient workflows for users.

This launch follows the introduction of GPT-6 Astra earlier this month, which is noted as OpenAI's most aligned model to date. The median daily token usage for researchers exceeds $600, with the 90th percentile surpassing $7,000, indicating a high demand for advanced AI capabilities in professional settings.