OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, expanding its GPT-6 family below Astra. Sol targets harder work such as coding. Luna is aimed at high-volume tasks such as extracting information and answering short questions. Both are available through the API, while access in OpenAI’s apps depends on the user’s plan and the ongoing rollout.
The company says API prices are 50% lower than the comparable GPT-5.6 Sol and Luna prices. OpenAI’s API changelog lists GPT-6 Sol at $2 per million input tokens, $0.20 per million cached input tokens, and $10 per million output tokens for standard requests with prompts up to 272,000 input tokens. GPT-6 Luna is $0.10, $0.01, and $0.50 respectively. Longer prompts and other processing tiers have different prices.
A lower token price does not settle model choice
OpenAI says Sol makes about half as many mistakes as its predecessor on an internal factuality evaluation based on conversations in which users flagged errors. That is a useful claim to test, but the sample comes from OpenAI’s own service and reflects the errors users noticed and reported. It does not establish the same reduction for a company’s documents, codebase, or customer requests.
The choice between Sol and Luna is especially consequential because their listed prices differ by a factor of 20 for fresh input and output tokens. A task that Luna completes reliably could cost far less. A task that needs repeated retries or human correction can erase that saving. Sol’s value depends on whether it avoids enough failures to justify its higher price on the work at hand.
OpenAI also compares the new models with Anthropic’s models, but its announcement notes that some competitor scores use earlier model versions when newer results were unavailable. Treat those comparisons as a starting point for testing rather than a ranking that applies to every workflow.
