OpenAI launches GPT-6 Sol and Luna at half the API price

OpenAI has launched GPT-6 Sol and GPT-6 Luna, two new models that extend the GPT-6 family below GPT-6 Astra. The announcement, published on September 22, 2026, describes Sol as the option for more demanding professional work and Luna as a faster, lower-cost model for higher-volume tasks. Both models use methods similar to Astra’s, with improvements aimed at coding, computer use, factuality, alignment and collaboration style. OpenAI says the models are designed to make advanced AI more practical at different scales and budgets.

The clearest commercial change is the API pricing. GPT-6 Sol is listed at $2 per million input tokens and $10 per million output tokens, down from $4 and $20 for GPT-5.6 Sol. GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, compared with $0.20 and $1.20 for GPT-5.6 Luna. OpenAI attributes the reductions to improved caching and inference, and says the new prices are 50% lower than the earlier promotional rates for the corresponding models.

OpenAI reports gains on several internal or provider-selected evaluations. Sol is presented as competitive on professional workflows, coding and computer use, while Luna improves on its predecessor at lower cost. The company also says Sol makes about half as many mistakes as GPT-5.6 Sol on its internal factuality evaluation. These figures are OpenAI’s own measurements and selected comparisons, not an independent guarantee for every task. The company notes that some competitor figures use public reports and that research-environment results can differ from production use.

Availability is broad but staged. GPT-6 Sol and Luna are available in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, and both are available through the API as gpt-6-sol and gpt-6-luna. Free and Go users can access Luna in the desktop app, while a gradual ChatGPT rollout was planned. GPT-6 Astra remains OpenAI’s highest-capability model.

For developers and businesses, the release lowers the cost of experimenting with coding agents, document workflows and computer-use systems. It may also encourage teams to route routine or high-volume tasks to cheaper models while reserving Astra for the hardest work. The practical test remains local: organisations must measure reliability, latency, permissions and total workflow cost before moving consequential tasks into production.