← All models

GPT-5.6 Luna

Published Publisher OpenAIgpt-5.6-luna

Overview

GPT-5.6 Luna is the fastest and most cost-efficient member of the GPT-5.6 family, designed for cost-sensitive, high-volume workloads. OpenAI describes it as roughly corresponding to the nano tier of earlier GPT-5 families. The family was released on July 9, 2026, following a limited preview. (developers.openai.com)

Capabilities and uses

  • Accepts text and image input and produces text output. It supports reasoning-effort settings from none through max, with medium as the default.
  • Offers a 1,050,000-token context window and a maximum of 128,000 output tokens. Its knowledge cutoff is February 16, 2026.
  • Supports streaming, function calling and structured outputs. Supported Responses API tools include web search, file search, code interpreter, computer use and MCP.
  • Suitable for high-volume applications, coding assistance, document and image analysis, and tool-using workflows where cost efficiency matters. (developers.openai.com)

Limitations

Audio and video are not supported, and fine-tuning is not supported. Model access and rate limits depend on eligibility and usage tier; a listed rate limit does not itself guarantee access. In OpenAI's evaluations, Luna generally trails the larger Sol model on demanding tasks and long-context retrieval; results depend on the evaluation and do not guarantee performance on a particular workload. (developers.openai.com)

OpenAI API pricing

USD per 1 million text tokens, checked November 16, 2026:

Category Price
Input $0.20
Cached input $0.02
Output $1.20

For prompts exceeding 272,000 input tokens, OpenAI charges the entire request at twice the input rate and 1.5 times the output rate. Cache writes are billed at 1.25 times the uncached input rate. Tool use can incur separate charges. (developers.openai.com)

The launch announcement originally listed Luna at $1 input and $6 output per 1 million tokens; its July 30, 2026 update announced an 80% price reduction. The prices in the table above are those shown on the model's official API page when checked, not the original launch prices. (openai.com)

Official sources