OpenAI expands GPT-6 lineup with lower-cost Sol and Luna models

OpenAI on Tuesday introduced GPT-6 Sol and GPT-6 Luna, lower-cost versions of its GPT-6 family, and began rolling them out across its developer tools and parts of ChatGPT. The new models sit below GPT-6 Astra, which OpenAI launched earlier this month as its flagship system.

The move broadens access to GPT-6 by offering much cheaper pricing while keeping the kind of very large context windows — the amount of text a model can take in at once — that matter for longer documents, coding sessions and repeated workflows. OpenAI said in a blog post, “Earlier this month, we introduced GPT‑6 Astra, the most intelligent and aligned model in the world. While the most demanding and important projects still call for Astra’s full depth, work happens at different scales, rhythms, and budgets.”

In OpenAI’s API, GPT-6 Sol is priced at $2.00 per 1 million input tokens and $10.00 per 1 million output tokens. GPT-6 Luna is priced far lower, at $0.10 per 1 million input tokens and $0.50 per 1 million output tokens. OpenAI said, “Prices are per 1 million tokens.” The company also said prompt-caching improvements now offer discounts of 90% on cached input-token reads, a pricing detail that could reduce costs for users running repeated prompts or longer-lived automated tasks.

OpenAI said GPT-6 Sol and GPT-6 Luna are available starting Tuesday in ChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu users. Free and Go users can access GPT-6 Luna in the desktop app. The models are also available in the API under the model names gpt-6-sol and gpt-6-luna. OpenAI said the new models are not yet available in Chat.

On technical specifications, both GPT-6 Sol and GPT-6 Luna have a 1,050,000-token context window and a maximum output of 128,000 tokens, according to OpenAI’s API documentation. The knowledge cutoff — the latest point in time reflected in a model’s training data — is listed as April 20, 2026, for GPT-6 Sol and May 18, 2026, for GPT-6 Luna.

OpenAI also published benchmark claims for the models, saying Sol and Luna performed well on tests including AutomationBench, Agents’ Last Exam, FrontierCode, DeepSWE and OSWorld. Those results are company-reported, not independently verified, and OpenAI said the evaluations were run in its research environment or through its API, which may differ from production ChatGPT output. On safety and behavior, OpenAI said the new models build on the alignment work introduced with Astra and improve on their GPT-5.6 counterparts, including showing lower rates of misleading claims about coding work.

The launch extends a tiered naming structure OpenAI had already used with GPT-5.6 in June, when it grouped models under Sol, Terra and Luna. With Tuesday’s announcement, OpenAI is carrying that approach into the GPT-6 line: Astra remains the top-end model, while Sol and especially Luna are designed to make newer GPT-6 capabilities available at lower cost.

That makes this less a brand-new GPT-6 debut than an expansion of the product range. For developers and workplace users, the main practical change is straightforward: OpenAI is pairing long context windows with sharply lower token prices, giving customers cheaper options for everyday AI tasks without moving off the GPT-6 platform.