GPT-6 Luna Review: Features, Benchmarks & Pricing

GPT-6 Luna is the smallest and cheapest model in OpenAI's GPT-6 family, released on September 22, 2026. It replaces GPT-5.6 Luna at the budget end of the lineup and sits below GPT-6 Sol, the mid-tier model that took over from GPT-5.6 Sol. OpenAI built it for work at scale: drafting, summarizing, sorting through long documents, and running the same kind of task thousands of times without the bill getting out of hand.

This GPT-6 Luna review covers what the model does well, what its benchmark numbers actually say, how much it costs, and where it falls short. It also covers how to use GPT-6 Luna without an OpenAI subscription — the model is already live in Chat Smith, with limited use on the free plan and higher usage limits on Chat Smith Pro.

GPT-6 Luna is OpenAI's efficient GPT-6 reasoning model for high-volume workloads, where speed and cost matter more than maximum depth. It accepts text and image input, returns text, and supports a context window of up to 1,050,000 tokens. Like the rest of the GPT-6 family, it lets you adjust reasoning effort, trading response speed for more careful answers on harder prompts.

OpenAI trained Luna with methods similar to those behind GPT-6 Astra, its flagship, so many of the same gains in professional work, factuality, coding, and computer use carry down to the cheaper tier. Compared with GPT-5.6 Luna, the new version writes better, operates software more reliably, and costs half as much per token.

It also reaches a much bigger audience than its predecessor. Since OpenAI's October update, GPT-6 Luna is the model behind GPT-6 in ChatGPT for Free and Go users, which makes it the GPT-6 model most people will actually use.

Four areas show the biggest change from the previous generation.

Luna inherits the communication style OpenAI introduced with Astra: clearer explanations, less jargon, fewer low-value details, and slightly shorter answers that keep the substance. For everyday writing — replies, summaries, product copy, or a quick draft from an AI email writer — that means less trimming before the text is ready to use.

On DeepSWE v1.1, which tests long-horizon software engineering in real codebases, GPT-6 Luna at max effort scored 66.6%. That is comparable to Claude Opus 5 and Claude Fable 5 at medium effort, at 93% and 96% lower cost per task respectively. OpenAI also reports fewer misleading claims about coding work than the GPT-5.6 models made, which matters when you let a low-cost model run unattended.

On OSWorld 2.0, a test of long computer-use workflows, Luna at max effort outscored GPT-5.6 Sol at medium effort for about a tenth of the cost. On AutomationBench, which runs business workflows across dozens of apps, Luna at high effort improved on GPT-5.6 Luna by 5.4 percentage points while cutting cost per task by 58%.

On OpenAI's internal factuality evaluation, built from real conversations where users flagged mistakes, GPT-6 Luna at higher effort levels matches GPT-5.6 Sol at roughly a hundredth of the cost. Combined with a 1.05M-token context window, that makes Luna a practical choice when you need to summarize a long PDF or pull answers out of a large set of notes without paying top-tier prices.

The table below compares GPT-6 Luna with the model it replaces and with GPT-6 Sol, the next tier up. Prices are OpenAI API rates per million tokens, and benchmark figures come from OpenAI's launch announcement.

The pattern is clear. Luna gives up a few points to Sol on hard coding and agentic tasks, but at one-twentieth of the token price. For most high-volume work, that trade favors Luna.

In the OpenAI API, GPT-6 Luna costs $0.10 per million input tokens and $0.50 per million output tokens, down from $0.20 and $1.20 for GPT-5.6 Luna. Cached input reads are discounted by 90%, which lowers costs further for agents and long conversations that reuse the same context.

• ChatGPT: powers GPT-6 in the Chat tab for Free and Go users, and is available in ChatGPT Work and Codex on Plus, Pro, Business, Enterprise, and Edu plans.
• Developers: available as gpt-6-luna in the OpenAI API, plus Microsoft Azure and Amazon Bedrock.
• GitHub Copilot: available on the Copilot Pro, Pro+, Max, Business, and Enterprise plans.
• Chat Smith: available on the free plan with limited usage, and with higher usage limits on Chat Smith Pro at $9.99 per month or $39.99 per year.

Luna is a budget model, and it behaves like one on the hardest problems. OpenAI still positions GPT-6 Astra for the most demanding work and GPT-6 Sol for complex everyday tasks, so long multi-step reasoning, tricky debugging, and high-stakes analysis are better sent to a bigger model.

Several of Luna's strongest results were measured at max effort. Higher effort means longer response times and more tokens per answer, so the real cost gap narrows when you push the model hard. At low effort, expect simpler output.

Benchmarks also do not map neatly onto your own workflow. Test GPT-6 Luna against other fast models such as Claude Haiku 4.5 or DeepSeek V4 Flash on your real tasks before you standardize on it.

You do not need an OpenAI account to use GPT-6 Luna. The model is already integrated into Chat Smith: free users get limited use, and Chat Smith Pro raises those limits for heavier work. Pick it from the model list and start chatting on Android, iOS, or in the browser.

The practical advantage is comparison. Luna sits next to other fast, low-cost models such as Gemini 3.1 Flash Lite in the same app, so you can send one prompt to several models and keep the answer that holds up. Context carries over when you switch models mid-conversation.

A sensible split in daily use: give GPT-6 Luna the routine, high-volume work — rewrites, summaries, quick research, first drafts — and move to a stronger model when a task needs deeper reasoning.

GPT-6 Luna brings a large share of the GPT-6 generation's gains to OpenAI's cheapest tier: better writing, stronger coding and computer use, more reliable facts, and a 1M-token context window, all at half the price of the model it replaces. It is not the model for your hardest problems, but for the bulk of everyday work it is hard to beat on cost.

If you want to test it yourself, GPT-6 Luna is already live in Chat Smith, a free AI chatbot that puts it next to the latest models from OpenAI, Anthropic, Google, xAI, and DeepSeek in one app.