OpenAI releases GPT-6 Sol and GPT-6 Luna

OpenAI released GPT-6 Sol and GPT-6 Luna, two cheaper, faster models that sit below its flagship GPT-6 Astra, released three weeks earlier, in a three-tier product line. Sol is pitched at coding and complex tasks; Luna, the cheapest of the three, at high-volume clerical work such as summarising documents or extracting information.

The company said that on its internal factuality evaluation — built from de-identified real-world conversations where users had flagged model mistakes — Sol makes roughly half as many errors as its predecessor, GPT-5.6 Sol, which it characterised as reaching “Astra-level reliability at much lower cost”. Pricing was cut alongside the accuracy claim: Sol now costs $2 per million input tokens and $10 per million output tokens, down from $4/$20 for GPT-5.6 Sol, while Luna costs $0.10/$0.50, down from $0.20/$1.20. OpenAI told VentureBeat the new prices were permanent rather than promotional.

Both models launched on 22 September in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users, and via the API as gpt-6-sol and gpt-6-luna. Free and Go users get Luna through the desktop app; neither model was yet available in the ordinary ChatGPT consumer tier at launch. As with GPT-6 Astra’s own release, the performance and error-rate figures are OpenAI’s own internal measurements rather than independently verified benchmark results.

The release filled out OpenAI’s pricing ladder beneath Astra with options aimed at cost-sensitive coding and high-volume text-processing workloads, continuing a pattern — also visible in rivals’ cheaper point releases — of frontier labs following a flagship launch with lower-cost variants aimed at capturing high-volume API traffic.