Developers can now deploy frontier-grade AI for professional automation at a fraction of the previous cost following the release of GPT-6 Sol and GPT-6 Luna from OpenAI. The new models are priced 50% lower than the GPT-5.6 family, with Luna costing $0.10 per million input and $0.50 per million output tokens, and Sol at $2 input and $10 output. Sol targets professional coding and matches performance on AutomationBench against Claude Opus 5 at only 9% of the cost per task, while Luna is designed for speed and high-volume scale. Both are currently live in the API and are rolling out to ChatGPT Work and Codex for Plus, Pro, Business, Enterprise, and Edu users.

The release comes as Anthropic also launched Claude Opus 5.5 to compete with the rising popularity of open-weight models. To further reduce operational expenses, OpenAI introduced improved prompt caching that offers discounts of up to 90% for reused input-token reads. This pricing offensive follows a period where leadership from both firms urged the industry to pace the frontier of AI development, shifting focus toward cost-per-task efficiency over raw capability jumps.