OpenAI: GPT-6 Luna

GPT-6 Luna is OpenAI's lowest-cost model in the GPT-6 family, released on September 22, 2026 alongside GPT-6 Sol. It is positioned for focused, high-volume workloads where teams need strong reasoning and tool use without paying flagship prices. The model is available only through OpenAI's API as gpt-6-luna and supports text and image inputs with text outputs, six reasoning-effort settings from none through max, and a 1.05 million-token context window with up to 128,000 output tokens.

Luna is trained with methods similar to GPT-6 Astra, bringing much of that generation's gains in professional work, factuality, coding, computer use, and alignment into a smaller, faster price tier. OpenAI reports competitive agent and coding results at max effort, including 66.6% on DeepSWE v1.1 and strong cost-per-task versus prior Luna and competitor models. Independent evaluations at max effort commonly report high scores on graduate-level Q&A and multimodal suites, while long-horizon office automation remains more challenging, reflecting Luna's efficiency-first design.

Standard API pricing is $0.10 per million input tokens and $0.50 per million output tokens, with aggressive cached-input discounts and batch or flex options for scale. Luna is available in ChatGPT Work, Codex, and the API for paid tiers, with limited desktop access for free and Go users. It does not ship open weights; deployment is fully hosted.