OpenAI released GPT-6 Sol and GPT-6 Luna on September 22, 2026, adding two models to the GPT-6 family it opened with GPT-6 Astra earlier in the month. According to the API changelog, both models accept text and image inputs and produce text through the Responses and Chat Completions APIs. A week later, on September 29, OpenAI followed with GPT-6.1 Sol, an upgrade to Sol.
On this page
Where the models are available
In its announcement, OpenAI says Plus, Pro, Business, Enterprise and Edu users can use both models in ChatGPT Work and Codex. Free and Go users get GPT-6 Luna only, in the ChatGPT desktop app. OpenAI notes the models are “not yet available in Chat,” and that the rollout would happen gradually over the launch day.
In the API, the model IDs are gpt-6-sol and gpt-6-luna.
OpenAI says Sol and Luna were trained with methods similar to GPT-6 Astra and bring the “advances behind Astra’s state-of-the-art performance in professional work, factuality, coding, computer use, and alignment” to faster models.
On September 25, the changelog recorded a fix for “a bug in image encoding that degraded image understanding in GPT-6 Sol and GPT-6 Luna,” affecting both the API and Codex.
Benchmark results
All scores below were published by OpenAI, with the reasoning effort OpenAI lists for each.
| Benchmark | Model | Score | Reasoning effort |
|---|---|---|---|
| DeepSWE v1.1 | GPT-6 Sol | 68.8% | max |
| DeepSWE v1.1 | GPT-6 Luna | 66.6% | max |
| Agents’ Last Exam | GPT-6 Sol | 56.4% | max |
| OSWorld 2.0 (offline) | GPT-6 Sol | 60.5% | xhigh |
| AutomationBench | GPT-6 Sol | 33.2% | xhigh |
OpenAI describes AutomationBench as testing agents on end-to-end workflows “using 47 tools across sales, marketing, operations, support, finance, and HR,” and Agents’ Last Exam as long-horizon tasks “spanning 55 sub-industries.” Lab-published scores for both models are tracked on our benchmark leaderboard.
GPT-6.1 Sol follows on September 29
OpenAI’s GPT-6.1 Sol announcement calls it “an upgrade to GPT‑6 Sol that nearly matches GPT‑6 Astra’s intelligence on agentic coding, computer use, and professional work.” The changelog lists it as built “for complex coding and professional work,” with text and code inputs. It is available to Plus, Pro, Business, Enterprise and Edu users in ChatGPT Work and Codex, and in the API as gpt-6.1-sol.
OpenAI reports that GPT-6.1 Sol matches GPT-6 Astra on DeepSWE v1.1, scoring 6.4 percentage points above GPT-6 Sol at maximum reasoning effort, and 7 points above GPT-6 Sol on OSWorld 2.0 at maximum effort.
What it changes
Paid ChatGPT users and API developers now have Sol, Luna and GPT-6.1 Sol alongside GPT-6 Astra, which OpenAI’s changelog calls its “most capable model.” For Free and Go users, Luna in the desktop app is the only one of these models available.
All four GPT-6 releases are listed on our AI model release timeline, and API service health is tracked on the OpenAI API status page.
