GPT-6 Sol, Luna cut API prices 50%

OpenAI introduced GPT-6 Astra earlier this month as its most intelligent and aligned model, and now it is expanding the family with GPT-6 Sol and GPT-6 Luna. The new models were trained with similar methods as Astra and carry over advances in professional work, factuality, coding, computer use, and alignment to faster, more affordable tiers. OpenAI says improvements in caching and inference let it serve these models at lower cost, and it is passing those savings on through the 50% API price reductions.

The company is positioning Sol and Luna as more cost-efficient options across the curve, with Sol exceeding Claude Fable 5.1 at far lower cost and even besting low-effort GPT-6 Astra on AutomationBench. Outside of raw capability, OpenAI also improved prompt caching for GPT-6, delivering higher cache hit rates by default and discounts of 90% on cached input-token reads, and it says GitHub reports the improvements reduced the share of prompt tokens requiring fresh processing by more than 50% across billions of requests. The alignment work from Astra also carries over, with both models showing improvements over their GPT-5.6 counterparts, including lower rates of misleading claims about their coding work.

The test for these models is likely to be whether the lower prices and caching gains translate into sustained agent use at scale, particularly for coding agents where OpenAI notes internal daily token usage has grown. Availability is starting in ChatGPT Work and Codex for paid users, with Free and Go users limited to GPT-6 Luna in the desktop app, so the breadth of adoption will hinge on how quickly the gradual rollout reaches users and whether developers shift workloads to the cheaper tiers.