AIToday
Large Language ModelsAI Business & IndustryLatent SpacePublished: Nov 18, 2026, 22:01 JST

GPT-6.2 Astra hits 62 as Sol ties old Astra at 15% cost

Send on LINE
GPT-6.2 Astra hits 62 as Sol ties old Astra at 15% cost

3 Key Points

  1. What happened

    OpenAI shipped GPT-6.2 Astra and Sol, rolling out to ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu and via API. Astra scored 62 on the Intelligence Index, just behind Fable 5.2 at 65, per Artificial Analysis.

  2. Why it matters

    Some early testers called Astra 'essentially unusable'.

WHO IT HITSEnterprise teams buying AI through ChatGPT Work, Codex or the API get cheaper per-task options — Astra at $8 input and $40 output per million tokens, Sol at $2 and $10 — but may need to weigh the mixed early tester feedback before switching workflows.

Not sure about something? Ask the AI

Questions and answers are published on this page.

Summaries like this, every morning.

Context & Analysis

OpenAI's launch lands two days before Anthropic's expected first trading day, and the timing shapes how the release reads: Anthropic still owns the world's best model, but OpenAI now owns the most efficient ones in terms of intelligence per dollar. Sol's ability to tie September's Astra at about 15% of the cost per task is the clearest expression of that shift, and the pricing table reinforces it — Astra at $8 input and $40 output per million tokens against Sol at $2 and $10.

The benchmark spread is wide and uneven. Astra leads Mercor's OpenWorldEngineer at 67.2% versus Fable 5.2 at 54.6%, and tops AA-3D at 1,318 ahead of Fable at 1,271, but trails Fable on the Intelligence Index. OpenAI also rated Astra High for AI Self-Improvement, a first for one of its externally released models.

What makes this launch unusual is the strongly negative influencer feedback. Some of it looks like the normal tendency to call everything either spectacular or the worst ever, but some criticisms appear valid, with noticeable and bizarre RL artefacts — and the launch video drew particular scorn. The counterweight is practical: the reviewers still use it to automate their work.

FAQ
How much do the new models cost?
Astra is $8 per million input tokens, $0.40 cached input and $40 output. Sol keeps its predecessor's pricing at $2, $0.10 and $10.
Where can I use them?
They are rolling out to ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu, and to developers through the API as gpt-6.2-astra and gpt-6.2-sol. Both have 1M context and a 128K output limit.
What did early testers say?
Matt Shumer called the models 'essentially unusable' after weeks of testing, while Pietro Schirano attributed many early issues to the harness.
Send on LINE

AI news that matters for your work, delivered every morning.

Pick your industry and the AI tools you use, and get news related to your work every day.

Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →

Ask AI

Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.

Questions and answers are published on this page.

Related Articles

Next articleSuno's Visuals adds AI video to songs via H3 Move