Start Chatting with GPT-5.6 Luna
Use GPT-5.6 Luna and its full model family, with more messages every day.
What is GPT-5.6 Luna?
GPT-5.6 Luna is OpenAI's cost-efficient model in the GPT-5.6 family. OpenAI positions it as the fastest and most cost-efficient GPT-5.6 model., designed for cost-sensitive, high-volume workloads and roughly corresponding to the nano model tier used in earlier GPT-5 families.
Released as part of the GPT-5.6 general availability launch on July 9, 2026, GPT-5.6 Luna is intended for developers and teams that need strong reasoning, coding, multimodal input, and tool-use support at a lower per-token cost than GPT-5.6 Terra or GPT-5.6 Sol.
GPT-5.6 Luna: Key Specs
Below are GPT-5.6 Luna's main specs and how they translate into real-world behavior.
- 1,050,000-token context window: GPT-5.6 Luna can handle very large prompts, long documents, repository context, or retrieval bundles, which helps reduce chunking in large-scale analysis workflows.
- 128,000 maximum output tokens: the model can generate long structured answers, summaries, extracted datasets, reports, and transformation outputs when a task requires extended text.
- Text and image input with text output: GPT-5.6 Luna can process written and visual context while returning text, supporting document review, screenshot analysis, multimodal classification, and visual data extraction.
- High reasoning with reasoning token support: GPT-5.6 Luna can spend reasoning tokens on harder tasks, making it more useful for complex classification, routing, coding, and analysis than a basic low-cost text model.
- $1.00 per 1M input tokens, $0.10 per 1M cached input tokens, and $6.00 per 1M output tokens: this pricing supports production workloads where cost per request matters.
- Streaming, function calling, structured outputs, and Responses API tools: GPT-5.6 Luna supports automation workflows that use web search, file search, image generation, code interpreter, hosted shell, apply patch, skills, computer use, MCP, and tool search.
Compare GPT-5.6 Luna, GPT-5.6 Sol & GPT-5.6 Terra
A brief overview of how each model differs in power, speed, and use cases.
| Feature | GPT-5.6 Luna | GPT-5.6 Sol | GPT-5.6 Terra |
|---|---|---|---|
| Knowledge Cutoff | February 16, 2026 | February 16, 2026 | February 16, 2026 |
| Context Window (Tokens) | 1,050,000 tokens | 1,050,000 tokens | 1,050,000 tokens |
| Max Output Tokens | 128,000 tokens | 128,000 tokens | 128,000 tokens |
| Input Modalities | Text, image | Text, image | Text, image |
| Output Modalities | Text | Text | Text |
| Latency (OpenRouter Data) | Fast; real-world latency varies by prompt length, output length, tool use, reasoning effort, and deployment conditions. | Fast; real-world latency varies by prompt length, output length, tool use, reasoning effort, and deployment conditions. | Fast; real-world latency varies by prompt length, output length, tool use, reasoning effort, and deployment conditions. |
| Speed | Fast | Fast | Fast |
| Input / Output Cost per 1M Tokens | $1.00 input, $0.10 cached input, $6.00 output | $5.00 input, $0.50 cached input, $30.00 output | $2.50 input, $0.25 cached input, $15.00 output |
| Reasoning Performance | High. Official GPT-5.6 Luna evaluations include GPQA Diamond 92.3%, FrontierMath Tier 1-3 v2 78.6%, FrontierMath Tier 4 v2 58.5%, and Artificial Analysis Intelligence Index v4.1 score 51.2. | Highest. Official GPT-5.6 evaluations list Sol at 52.7% on Agents' Last Exam, 1,747.8 Elo on GDPval-AA v2, 58.9 on Artificial Analysis Intelligence Index v4.1, 90.4% on BrowseComp, and 62.6% on OSWorld 2.0. | High. Official GPT-5.6 evaluations list Terra at 50.4% on Agents' Last Exam, 55 on Artificial Analysis Intelligence Index v4.1, 57.7% on HealthBench Professional, and 87.5% on BrowseComp. |
| Coding Performance (on SWE-bench Verified) | Strong for a lower-cost model. Official evaluations include Artificial Analysis Coding Agent Index v1.1 score 74.6, SWE-Bench Pro 62.7%, DeepSWE v1.1 67.2%, and Terminal-Bench 2.1 84.7%. | Very strong. Official GPT-5.6 evaluations list Sol at 80 on Artificial Analysis Coding Agent Index v1.1, 64.6% on SWE-Bench Pro, 72.7% on DeepSWE v1.1, and 88.8% on Terminal-Bench 2.1. | Strong. Official GPT-5.6 evaluations list Terra at 77.4 on Artificial Analysis Coding Agent Index v1.1, 63.4% on SWE-Bench Pro, 69.6% on DeepSWE v1.1, and 87.4% on Terminal-Bench 2.1. |
| Best For | High-volume classification, summarization, routing, structured extraction, real-time applications, coding support, and cost-sensitive agent workflows | Complex reasoning, professional coding, long-context research, knowledge work, cybersecurity, science, computer use, design, and advanced agent workflows | Production coding, structured data extraction, content workflows, general-purpose agents, long-context analysis, and cost-aware knowledge work |
Best Cases to Use GPT-5.6 Luna
Best GPT-5.6 Luna use cases
- high-volume classification systems that need fast, cost-efficient decisions across text, documents, or image-based inputs.
- summarization and routing workflows where the model condenses, labels, prioritizes, or sends work to other models or tools.
- structured data extraction from support tickets, forms, PDFs, screenshots, long documents, and retrieval-augmented context bundles.
- real-time applications where lower latency and lower per-token cost matter more than using the highest-capability model in the GPT-5.6 family.
- coding support and agent substeps such as repository search, issue triage, simple code edits, test interpretation, and tool-coordinated workflows.
- education, writing, research, and productivity assistants that need long context, image input, and reliable text output without flagship pricing.
How to Access GPT-5.6 Luna
How to access GPT-5.6 Luna
1. Official API
OpenAI lists GPT-5.6 Luna as gpt-5.6-luna in the API. It supports the Chat Completions and Responses endpoints, text and image input, text output, streaming, function calling, structured outputs, and supported Responses API tools.
2. EssayDone AI Chat
EssayDone AI Chat provides access to this model through an easy-to-use chat interface when the model is available in the product.
This can be useful for writing, study, research, summarization, extraction, and productivity tasks, but it does not replace the official API for developers who need direct control over endpoints, parameters, tools, and deployment settings.
Explore More AI Models
Find the model you need-search or select to open its full profile.
19 models available
FAQ
Here are some frequently asked questions about GPT-5.6 Luna.
Yes. GPT-5.6 Luna supports reasoning tokens and is listed by OpenAI with High reasoning capability. It is designed to bring strong reasoning to cost-sensitive workloads rather than to replace the higher-capability GPT-5.6 Sol model.
OpenAI lists GPT-5.6 Luna API pricing at $1.00 per 1M input tokens, $0.10 per 1M cached input tokens, and $6.00 per 1M output tokens. Prompts with more than 272K input tokens are priced at 2x input and 1.5x output for the full request, and cache writes are billed at 1.25x the uncached input token rate.
GPT-5.6 Luna is optimized for cost-sensitive, high-volume work such as classification, summarization, routing, structured extraction, real-time applications, lightweight coding support, and agent workflow substeps.
GPT-5.6 Luna supports text and image inputs and returns text output. It can handle multimodal workflows such as screenshot analysis, document review, image-aware classification, and visual information extraction, but audio and video are not supported.
Compared with GPT-5.6 Terra and GPT-5.6 Sol, GPT-5.6 Luna is the fastest and lowest-cost model in the family. OpenAI describes it as roughly corresponding to the earlier nano tier, while still offering GPT-5.6-family long context, reasoning, coding, and tool-use capabilities.
Using GPT-5.6 Luna in EssayDone AI Chat can provide a convenient way to use a fast, cost-efficient OpenAI model for writing, research, summarization, study, extraction, and everyday productivity tasks without setting up the API.