OpenAI launches GPT-6 Sol and Luna at half the API price
Original: OpenAI launches GPT-6 Sol and Luna, boasting lower cost and fewer mistakes
Why This Matters
A 50% API price cut signals intensifying cost competition at the frontier model tier.
OpenAI on September 22, 2026 released updated GPT-6 Sol and Luna models, offering API access at half the cost of the 5.6 series. Sol targets complex coding tasks; Luna handles high-volume clerical work. GPT-6 Sol reportedly cuts factual errors by 50% versus its predecessor, matching Astra-level accuracy at lower cost.
Just weeks after launching GPT-6 Astra — billed as its most capable model — OpenAI is broadening the GPT-6 lineup with refreshed versions of Sol and Luna, two smaller models occupying different rungs of its model hierarchy. Sol is built for demanding tasks like coding; Luna targets lighter, high-volume work such as document summarization and quick Q&A.
The headline figure is price: API access for the new Sol and Luna comes at half the cost of the previous 5.6 series, a reduction OpenAI attributes to gains in caching and inference efficiency. On accuracy, OpenAI says GPT-6 Sol makes roughly half as many factual errors as its predecessor, based on internal evaluations drawn from real-world conversations where users flagged model mistakes — putting it at 'Astra-level reliability at much lower cost,' per the announcement.
Competitive timing was conspicuous. Anthropic dropped a new version of Opus 5.5 just 90 minutes before OpenAI's release. OpenAI, in turn, claims GPT-6 Sol and Luna outperform Anthropic's top models, including Fable and Opus, across a range of tasks.
GPT-6 Sol is now live in ChatGPT Work, Codex, and the API for most paid users. Luna is rolling out to the desktop app and Free and Go tiers as well, with broader ChatGPT availability expected throughout the day.