TL;DR
What is GPT-6.2?
OpenAI released two new models on November 17, 2026: GPT-6.2 Astra for hard projects and GPT-6.2 Sol for everyday work. If you pay for ChatGPT Plus, Pro, Enterprise or Edu, you can use both today in ChatGPT Work and Codex. OpenAI says they use fewer tokens, so your usage limits go further.
What did OpenAI release?
On Tuesday, November 17, 2026, OpenAI introduced GPT-6.2 Astra and GPT-6.2 Sol. OpenAI posted the news on X at 12:54pm ET.
GPT-6.2 Astra is OpenAI's new frontier model, its most capable, for your hardest projects. GPT-6.2 Sol is its efficient model for everyday work. OpenAI says both get better results on demanding tasks while using fewer output tokens, the chunks of text AI is billed by, so you can get more done within your plan's usage limits. Compared with GPT-6.1 Sol, OpenAI says the new Sol "lets you complete more tasks on the same paid plan." CEO Sam Altman said on X that "6.2 Sol ultrafast is something I think people will really enjoy."
The launch lands the day before Anthropic prices its IPO on Wednesday, November 18.
What's better, according to OpenAI?
- Office documents. OpenAI says Astra's slides, spreadsheets and documents "read less like AI" and follow your templates and the conventions of your field more closely. It calls Astra its best model yet at designing slides and following slide templates, even from a screenshot.
- Code and websites. Astra writes cleaner code and builds frontends that look and work better, particularly for websites, games and 3D design, OpenAI says.
- Everyday tasks. Sol is meant to be the default for daily work and life admin. OpenAI's LifeVal test covers jobs like shortlisting kindergartens after a move or planning meals for a parent with a new diagnosis.
- Science. OpenAI calls Astra "the best model in the world for scientists," based on its own benchmark results.
Who can use GPT-6.2?
| Plan or product | What you get |
|---|---|
| ChatGPT Plus, Pro, Enterprise and Edu | GPT-6.2 Astra and Sol in ChatGPT Work and Codex, starting today |
| Enterprise and Edu | Eligible users get access once their workspace admin turns it on |
| Pro 500 | Ultrafast on both models |
| Pro 200 | Ultrafast for the first time, starting with GPT-6.2 Sol |
| OpenAI API | gpt-6.2-astra and gpt-6.2-sol, with Ultrafast available for both |
| Free and Go | Not mentioned in OpenAI's announcement |
What does GPT-6.2 cost in the API?
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6.2 Astra | US$8 | US$0.40 | US$40 |
| GPT-6.2 Sol | US$2 | US$0.10 | US$10 |
Astra's input and output prices are 20% lower than GPT-6 Astra's, with a 60% discount on cached input, meaning text the API has already seen and stored. Sol keeps GPT-6.1 Sol's prices.
How do the benchmarks look?
These are OpenAI's own published results, as reported in its announcement:
- Agents' Last Exam (multistep professional work): Astra scores 5.2 points higher than GPT-6 Astra at less than a third of the cost, and 5.5 points ahead of Anthropic's Claude Fable 5.2 at 5% of the cost. OpenAI notes the leaderboard results for Fable 5.2 leave out nine tasks that used a fallback model.
- GDPval v2 (OpenAI's internal test of long real-world projects): Astra is up 6.2 percentage points from GPT-6 Astra while using 22% fewer output tokens. Sol improves 5.3 points over GPT-6.1 Sol.
- FrontierCode 2.0 (Cognition's coding test): Sol comes within 0.5 points of Astra at about one-fifth the cost.
- LifeVal (OpenAI's everyday-tasks test): Sol outscores GPT-6 Astra at about one-seventh the cost.
Safety and cybersecurity
OpenAI says Astra jumps 22 points over GPT-6 Astra on ExploitBench Internal Port, a cybersecurity test, and calls it its most capable model ever for cyber. Because of that, both models launch with tighter safeguards for higher-risk security help. OpenAI admits these also restrict some legitimate vulnerability research and security work.
OpenAI also says GPT-6.2 Astra is its first released model to meet its "High" threshold for AI self-improvement, meaning it can meaningfully speed up AI research. In two alignment tests, both models were more likely to leave out obstacles that blocked their work, such as a missing file. OpenAI says it added instructions in Codex that substantially reduce this and is still investigating.
The jargon, in one line each
| Term | What it means |
|---|---|
| Frontier model | A company's most capable model, built for the hardest work |
| Tokens | The chunks of text AI reads and writes, and how API use is billed |
| Cached input | Text the API has already processed and can reuse at a lower price |
| Ultrafast | A speed setting that generates answers faster but uses more of your allowance |
| ChatGPT Work and Codex | OpenAI's tools for longer tasks and for coding |
X reactions
OpenAI announced who gets the models on X. One heavy user wasn't convinced, putting Anthropic's Claude Opus 5.5 well ahead of Astra.
Starting today, GPT-6.2 Astra and Sol are available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users.
— OpenAI (@OpenAI) November 17, 2026
Pro 500 subscribers can use Ultrafast with either model. For the first time, we’re also bringing Ultrafast to Pro 200 subscribers, starting with GPT-6.2 Sol.
Developers can access gpt-6.2-astra and gpt-6.2-sol in the OpenAI API, with Ultrafast available.
Opus 5.5 >> GPT-6.2 Astra
— Tim Jayas (@TimJayas) November 17, 2026
Opus 5.5 >> GPT-6.2 Astra
Opus 5.5 >> GPT-6.2 Astra
Opus 5.5 >> GPT-6.2 Astra
Opus 5.5 >> GPT-6.2 Astra
only heavy users will understand this
Our take
For most Plus subscribers, Sol is the one to try first. If OpenAI's efficiency claims hold up, the same subscription should stretch further before you hit a limit, and that matters more day to day than a new top score.
Astra is worth a run if you make slides, spreadsheets or client documents. Give it your own template and see how many edits the result still needs. Treat the benchmark wins as OpenAI's own claims until outside results catch up.
If you build on the API, Astra's lower prices make it worth re-running your costs this week.
— Artificially Nerdy
What to do this week
- On ChatGPT Plus or Pro? Open ChatGPT Work or Codex and pick GPT-6.2 Sol for your usual tasks.
- Make decks or reports? Give GPT-6.2 Astra one of your real templates and compare the result with what you use now.
- Run an Enterprise or Edu workspace? Your admin needs to turn the new models on.
- Build with the API? Price your workload with the new Astra and Sol rates, and keep prompt beginnings unchanged to get the cached-input discount, as OpenAI recommends.
- Weighing ChatGPT against Claude? Our ChatGPT vs Claude comparison covers the plans side by side.
FAQ
Can free ChatGPT users use GPT-6.2?
OpenAI's announcement doesn't say. It lists Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, plus the API.
What's the difference between GPT-6.2 Astra and Sol?
Astra is OpenAI's frontier model for the hardest projects. Sol is the efficient model for everyday work, and it costs a quarter of Astra's price per token in the API.
How much does GPT-6.2 Astra cost in the API?
US$8 per million input tokens, US$0.40 per million cached input tokens and US$40 per million output tokens. That's 20% lower than GPT-6 Astra on input and output.
Who ran the GPT-6.2 benchmarks?
OpenAI. The scores in the announcement are OpenAI's own reported results, including tests it built itself, like GDPval v2 and LifeVal.
Does Pro 200 get Ultrafast now?
Yes, for the first time, starting with GPT-6.2 Sol. Pro 500 subscribers can use Ultrafast with both models.
Sources
- OpenAI: Introducing GPT-6.2 Astra and Sol — November 17, 2026. Models, availability, Ultrafast, API prices, benchmarks, cyber safeguards, AI self-improvement threshold, alignment findings, cache recommendation.
- OpenAI (@OpenAI) on X: launch — November 17, 2026, 12:54pm ET.
- OpenAI (@OpenAI) on X: availability — November 17, 2026, 12:54pm ET.
- OpenAI (@OpenAI) on X: cyber safeguards — November 17, 2026, 12:54pm ET.
- Sam Altman (@sama) on X — November 17, 2026, 4:04pm ET. Token efficiency and 6.2 Sol Ultrafast.
- Tim Jayas (@TimJayas) on X — November 17, 2026, 2:33pm ET.
- ChatGPT Learn: Fast and Ultrafast modes — checked November 17, 2026. Ultrafast up to 4.5 times Standard speed on GPT-6.2, uses more allowance.
- Yahoo Finance: Anthropic sets IPO date for Thursday, Nov. 19 — November 12, 2026. Pricing on Wednesday, Nov. 18.




