On September 22, OpenAI released GPT-6 Sol. On September 29, it released GPT-6.1 Sol, and the GPT-6 Sol page now points readers to the newer model. That is one week as the current Sol. API changelog.
Both list at $2 input / $10 output per million tokens. The pricing difference is cached input, which falls from $0.20 to $0.10 on GPT-6.1 Sol. OpenAI positions the model as near-Astra performance for complex coding, computer use, and professional work. Astra costs $10 / $50, so Sol's standard rates are one-fifth of Astra's. Pricing.
The API model is gpt-6.1-sol. It takes text and images, returns text, and has a 1,050,000-token context with up to 128,000 output tokens. Documented reasoning effort runs from low through max, with medium as the default. Tool calling requires the Responses API, where multi-agent runs are available in beta. Model page.
The long-context rule still applies. Above 272K input tokens, the request is billed at 2× input and 1.5× output. Our context calculator includes that threshold.
GPT-6 Luna shipped in the September 22 release and remains current. It costs $0.10 / $0.50 per million tokens, half of GPT-5.6 Luna, and supports effort from none through max. One catch from its model page: in Chat Completions, function calling works only with reasoning effort set to none.
OpenAI's model guide now recommends three models: Astra for the hardest work, GPT-6.1 Sol to balance capability and cost, and GPT-6 Luna for high-volume tasks. There is no GPT-6 Terra. The GPT-5.6 tiers are still on the pricing page, and the comparison is awkward for them. GPT-5.6 Sol is $4 / $20, a rate OpenAI calls promotional through at least November 21. GPT-5.6 Terra is $2 / $12. The newer Sol lists below both.
A lower list price does not settle the cost of a finished task. Effort level, output length, and retries decide that.
The same day, OpenAI added an Ultrafast processing tier for Astra, selected with service_tier: "ultrafast". Check the pricing page for the premium before enabling it on a long run.
GPT-6.1 Sol and GPT-6 Luna are now in the model tools, both marked “not tested.” GPT-6 Sol is not, since it was superseded within a week. The GPT-5.6 results stay under their own names. If GPT-5.6 Sol is your default today, replay one long task on 6.1 Sol through the Responses API with the same acceptance tests, and compare total spend.