GPT-6 Luna

Released 2026-09-22 per OpenAI's own API changelog ("Released GPT-6 Sol (gpt-6-sol) and GPT-6 Luna (gpt-6-luna). These reasoning models accept text and image inputs and generate text", tagged v1/responses and v1/chat/completions). The model page calls it "our most efficient model for focused, high-volume tasks". Text+image input, text output; 1,050,000-token context, 128,000 max output. reasoning.effort supports none/low/medium (default)/high/xhigh/max; use the Responses API for built-in tools and function calling - Chat Completions supports function calling only with reasoning_effort none. Cached input is 10% of uncached input and cache writes are 1.25x it. Prompts >272K input tokens are billed at 2x input and cache rates and 1.5x output for the full request; the price fields carry the <=272K band. Batch and Flex are 50% of Standard; Fast mode is 2x the applicable rates; regional processing adds a 10% premium where available, and EU data residency is available only with Standard processing. Standard tier per 1M: $0.10 in / $0.01 cached in / $0.125 cache write / $0.50 out, stated identically on the model page, the pricing page and the changelog entry. May 18 2026 knowledge cutoff. OpenAI states no relationship between this model and gpt-5.6-luna - no deprecation of that row, no migration pointer - so none is recorded. STATUS: recorded as ga. OpenAI's changelog says 'Released' and the model page carries no access qualifier - unlike GPT-6 Astra's launch page on 2026-09-03, which carried a staged Trusted Access rollout sentence and was held at preview for that reason. OpenAI does not use the words 'generally available' here either, so this is the same positive-content reading the Astra row settled on, not a quoted GA. Lifecycle fields are null: OpenAI states no deprecation for this model.

Source: OpenAI official documentation ↗

Explore more

GPT-6 Luna — questions & answers