San Francisco, 22 September 2026 — OpenAI has released two new models in its GPT-6 generation, GPT-6 Sol and GPT-6 Luna, aimed at making its most capable technology cheaper and more widely available, TechCrunch reported.

The models follow GPT-6 Astra, which OpenAI launched on 3 September 2026. OpenAI said Sol and Luna "extend its benefits by making that intelligence more efficient and accessible."

OpenAI claims GPT-6 Sol makes about half as many mistakes as its predecessor and reaches Astra-level reliability, with better factual accuracy and lower coding error rates. OpenAI also claims the models outperform rival systems on its benchmarks. These claims come from the company and have not yet been independently verified.

The split between a heavyweight model and a fast, cheap model reflects a wider industry trend: businesses increasingly route tasks to different models depending on complexity and cost. Lower prices could also accelerate AI adoption by startups and developers, including in cost-sensitive markets such as India.

The rollout to ChatGPT users was expected to continue gradually. Developers and independent researchers will now test the models against OpenAI's claims on accuracy, coding and cost.

Source: TechCrunch (22 September 2026); OpenAI announcements.