GPT-6 Luna
OpenAI · Multimodal model
GPT-6 Luna is OpenAI's small, low-cost model in the GPT-6 family, released alongside GPT-6 Sol. It targets focused, high-volume work where price and speed matter more than peak capability.
Current
ReleasedSep 22, 2026
Context window1.05M tokens (922K max input, 128K max output)
Input → outputtext, image → text
API price$0.10 / 1M input tokens, $0.01 / 1M cached input tokens, $0.50 / 1M output tokens
Consumer accessChatGPT Work and Codex for Plus, Pro, Business, Enterprise and Edu; Free and Go users can try it in the ChatGPT desktop app
Best for
High-volume, latency- or cost-sensitive tasks such as classification, extraction and simple agent steps.
Overview
GPT-6 Luna is OpenAI's small, low-cost model in the GPT-6 family, released alongside GPT-6 Sol. It targets focused, high-volume work where price and speed matter more than peak capability.
Strengths and limits
- OpenAI's cheapest and fastest GPT-6 model
- Supports 'none' reasoning effort for low-latency calls
- Same 1M+ token context window as Astra and Sol
- Much weaker than Sol/Astra on agentic coding in third-party comparisons
- Function calling in Chat Completions only works with reasoning effort 'none'
Published benchmark scores
Terminal-Bench 4.016.4%
Source: www.anthropic.com
OSWorld 2.1 (offline subset)48.9%
Source: www.anthropic.com
GDPval-AA v2.11437
Source: www.anthropic.com
Scores are as published by the source; settings (reasoning effort, tools, attempts) differ between labs, so compare with care. How to read benchmarks →