AI Potluck
Back to Gap Map Model components / Fine-tuned / chat models

GPT-6

OpenAI
closed / Overall score: n/a

GPT-6 is OpenAI's flagship model family, released in September 2026 in three tiers: Astra, the most capable, for complex reasoning, coding and computer use; Sol, which balances capability and cost and has been updated as GPT-6.1 Sol; and Luna, for high-volume work. All three take text and image input with a context window of about one million tokens. They are served through OpenAI's API and ChatGPT, without downloadable weights.

Openness

1 high confidence
1.0
weights
closed
data
closed
code
closed
license
Proprietary(API + ChatGPT only)

GPT-6 has no downloadable weights, no published training data and no published training code. Every tier is served only through OpenAI's API and ChatGPT and billed per token.

  • https://developers.openai.com/api/docs/models/gpt-6-astra recorded 2026-10-09

    OpenAI API model page: "GPT-6 Astra is our most capable model for the most demanding work", priced per token ("Pricing is based on the number of tokens used"). Searched the body for weights, open-weight, download, Hugging Face and license: none; the model is offered only as a hosted API endpoint.

  • https://developers.openai.com/api/docs/models/gpt-6-luna recorded 2026-10-09

    OpenAI API model page: "GPT-6 Luna is our most efficient model for focused, high-volume tasks", priced per token. Searched the body for weights, open-weight, download, Hugging Face and license: none.

  • https://en.wikipedia.org/wiki/GPT-6 recorded 2026-10-09

    Wikipedia GPT-6 article: infobox license "Proprietary"; GPT-6 is "a family of large language models (LLM) developed by American artificial intelligence firm OpenAI", with Astra "released to the general public on September 4, 2026" and Sol and Luna on September 22, 2026. Searched the body for open weights, training data release, published code and open source: none.

Adoption

not assessed

OpenAI has published no usage count for GPT-6. ChatGPT's earlier user figures predate the family, and GPT-6 only became the default ChatGPT model in October 2026, so no level is assigned yet.

Capability

5 high confidence
5.0

This score reads GPT-6 Astra, OpenAI's most capable model. On the Artificial Analysis index it sits in the leading group with the Claude Opus and Fable models, level with Claude Fable 5.1 and about five points under Claude Opus 5.5, and the cheaper GPT-6.1 Sol scores almost as high.

  • https://artificialanalysis.ai/leaderboards/models recorded 2026-10-09

    The leaderboard's embedded model records, ranked by intelligenceIndex, put Claude Opus 5.5 (Max) first at 57.62, then Claude Fable 5.1 (Max) 53.35, GPT-6 Astra (Max) 52.67 and GPT-6.1 Sol (Max) 51.83 among models not marked deprecated; Grok 4.7 (Xhigh) follows at 46.45. The digest drifts between fetches because the page carries live pricing and throughput figures.

  • https://artificialanalysis.ai/models/gpt-6-1-sol recorded 2026-10-09

    "GPT-6.1 Sol (Max) scores 52 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models (median: 26)"; Intelligence rank "# 11 / 226".

  • https://artificialanalysis.ai/models/gpt-6-astra recorded 2026-10-09

    "GPT-6 Astra (Max) scores 53 on the Artificial Analysis Intelligence Index, placing it well above average among comparable models (median: 26)"; Intelligence rank "# 7 / 226".

  • https://developers.openai.com/api/docs/models/gpt-6-astra recorded 2026-10-09

    OpenAI API model page: GPT-6 Astra is "our most capable model for the most demanding work. Use it for complex reasoning, coding, computer use, research, and document creation", with a "1,050,000 context window"; the page carries no benchmark figures.

Verified 2026-10-09, except some scores, still awaiting re-confirmation