OpenAI is expanding its GPT-6 family with Sol and Luna. Both models are available starting today in ChatGPT Work, Codex, and the API. API prices are dropping by 50 percent compared to the GPT-5.6 variants. GPT-6 Astra remains the company’s most powerful model.
With Sol and Luna, OpenAI is explicitly targeting the cost-sensitive segment of the market. According to the company, the two models were trained using the same methods as Astra but run faster and cost less. Improvements in caching and inference make this possible. The company passes the savings directly on to customers.
The price tags are significantly lower. GPT-6 Sol costs $2 per million input tokens and $10 per million output tokens, whereas GPT-5.6 Sol was still priced at $4 and $20, respectively. Luna drops from $0.20 to $0.10 per million input tokens and from $1.20 to $0.50 per million output tokens. By comparison, Astra charges $10 for input and $50 for output per million tokens.
Benchmarks against Claude
OpenAI primarily compares the new models against Anthropic. On AutomationBench, Sol scores 33.2 percent on xhigh effort, compared to 26.9 percent for Claude Opus 5, at one-ninth the cost per task. On DeepSWE, Sol achieves 68.8 percent, slightly below Claude Fable 5’s 69.9 percent, but at about 80 percent lower cost. Luna achieves 66.6 percent there.
OpenAI also claims progress in factual accuracy: according to internal tests, Sol makes about half as many errors as its predecessor.
Caching and availability
For developers, there’s more than just a lower token price. Prompt caching has been improved, with a 90 percent discount on cached input tokens. GitHub reports that as a result, the proportion of prompt tokens that need to be reprocessed has dropped by more than half.
Sol and Luna are available starting today for Plus, Pro, Business, Enterprise, and Edu users. Free and Go users will get Luna in the desktop app. The models aren’t available in Chat yet.
Tip: Claude Opus 5.5 released: better and cheaper than Fable 5.1