GPT-6 explained: Astra, Sol and Luna models

The four GPT-6 models

• GPT-6 Astra: OpenAI's most capable model, for complex reasoning, coding, computer use, research and document creation. $10 input, $1 cached input, $50 output per million tokens. See our GPT-6 Astra page.
• GPT-6.1 Sol: "near-Astra performance" at a lower cost, for complex coding, computer use and professional work. $2 input, $0.10 cached input, $10 output.
• GPT-6 Sol: built for complex coding and agentic workflows. OpenAI points to GPT-6.1 Sol as the newer Sol model. $2 input, $0.20 cached input, $10 output.
• GPT-6 Luna: OpenAI's most efficient model for focused, high-volume tasks. $0.10 input, $0.01 cached input, $0.50 output.

Prices are OpenAI's standard API rates for prompts up to 272K input tokens, from each model page. Longer prompts are billed at 2x the input and cache rates and 1.5x the output rate for the whole request. Batch and Flex are 50% cheaper, and Fast mode costs 2x.

GPT-6 in ChatGPT's Chat tab

On 7 October 2026 OpenAI began rolling out GPT-6 with Intelligent UI in the Chat tab, starting with Plus, Pro, Business and Enterprise and expanding to Free and Go the next day. OpenAI says both Chat models are tuned for everyday conversation, and that the update doesn't change the models behind Work and Codex. Separately, OpenAI's Codex docs list GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna as available in Work and Codex, not in Chat.

GPT-6 in Codex

• OpenAI recommends GPT-6.1 Sol for complex coding and agentic work, Luna for focused, repeatable tasks, and Astra when a task needs the strongest capability.
• Free and Go plans get GPT-6 Luna in the desktop app, subject to rollout.
• Pick a model with codex --model gpt-6.1-sol, with /model in the CLI, or with model = "gpt-6.1-sol" in config.toml.
• GPT-5.5 retires from ChatGPT, Work and Codex on 14 October 2026. OpenAI's replacements are gpt-6-sol on paid plans and gpt-6-luna on Free and Go. The API isn't affected.

Reasoning effort across the family

In the API, GPT-6 Astra and GPT-6.1 Sol support low, medium, high, xhigh and max, but not none. GPT-6 Sol and GPT-6 Luna also support none. GPT-6.1 Sol, GPT-6 Sol and GPT-6 Luna default to medium, per OpenAI's reasoning guide. Tool calling with Astra and GPT-6.1 Sol requires the Responses API.

Independent benchmarks

Artificial Analysis reported a score of 53 for GPT-6 Astra on its Intelligence Index (13 Oct 2026), the highest of the family. Results for all four models from open datasets (Arena, Epoch AI) are on our benchmarks page.