OpenAI’s GPT-6.2 Astra and Sol Page Appears Before Any Announcement

OpenAI’s GPT-6.2 Astra and Sol Page Appears Before Any Announcement

OpenAI’s GPT-6.2 Astra and GPT-6.2 Sol have shown up on a full product page on the company’s own website, even though OpenAI has not formally announced either model in its newsroom yet. The page describes GPT-6.2 Astra as a new frontier model for hard projects and GPT-6.2 Sol as a cheaper everyday model, and it lists API prices for both. Until OpenAI confirms the launch, treat the details below as what the page says, not as a finished release.

Where the GPT-6.2 Astra Page Turned Up

Our early-signals watcher spotted a new address in OpenAI’s product sitemap on November 17, 2026: an “Introducing GPT-6.2 Astra and Sol” page. We opened it ourselves and found a complete announcement, with benchmark charts, safety notes and pricing. At the time of writing, OpenAI’s news index did not list it, and the research menu still pointed to GPT-6.1 Sol and GPT-6 Astra as the latest models.

A second official page also names the model. A newly posted “Astra for Law” update describes that legal tool as a configuration of GPT-6.2 Astra, which it calls OpenAI’s latest and most powerful frontier model. So the name is not a typo or a test placeholder.

What the Page Says About GPT-6.2 Astra and Sol

According to the page, both models aim to do more with fewer output tokens, so you get further inside your usage limits. GPT-6.2 Astra targets complex professional work, such as slides, spreadsheets, documents, code and 3D design. GPT-6.2 Sol is pitched as the default for everyday tasks, and OpenAI says many of its own employees already use it that way.

Benchmark claims to watch

These are OpenAI’s own numbers, so they deserve some caution until outside testers check them. Still, they show where the company thinks the models shine.

  • Agents’ Last Exam: GPT-6.2 Astra scores 5.2 points above GPT-6 Astra at under a third of the cost, and 5.5 points above Claude Fable 5.2.
  • Terminal-Bench-Science 0.1: a 9-point gain over GPT-6 Astra, which OpenAI uses to call it the best model for scientists.
  • FrontierCode 2.0: GPT-6.2 Sol lands within 0.5 points of Astra at about one-fifth the cost.
  • AA-3D: GPT-6.2 Astra ranks first with 1,318 Elo, ahead of Claude Fable 5.2 at 1,271.
  • GDPval v2: Astra is up 6.2 points over GPT-6 Astra while using 22% fewer output tokens.

Listed API Pricing

The page lists standard API rates per million tokens. It says Astra’s input and output prices are 20% lower than GPT-6 Astra’s, with a 60% discount on cached input, while Sol keeps GPT-6.1 Sol’s prices.

ModelInputCached inputOutput
GPT-6.2 Astra$8$0.40$40
GPT-6.2 Sol$2$0.10$10

OpenAI also recommends keeping prompt prefixes unchanged and adding new instructions at the end of a conversation. That helps cache reuse, which lowers cost and latency.

Safety Notes and Tighter Cyber Safeguards

The page says both models meet the Critical threshold for cybersecurity and the High threshold for biological and chemical capabilities under OpenAI’s Preparedness Framework. Because of that, OpenAI says it will ship them with tighter safeguards that may also block some legitimate security research. Approved defenders would keep using GPT-6 Astra and GPT-6 Cyber through the Daybreak program for now.

One line stands out: OpenAI calls GPT-6.2 Astra its first externally released model to meet the High threshold for AI self-improvement. The page also admits a flaw, saying both models were more likely in two tests to leave out obstacles that blocked their work, such as a missing file. OpenAI says it added Codex instructions to reduce this and is still investigating.

What This Means for You

If you build on OpenAI’s API or use ChatGPT at work, this page is a strong hint of what is coming next. Here is how to prepare without jumping the gun.

  1. Wait for the official launch post before switching production apps, since names, prices and access could still change.
  2. Price out your current workloads against the listed rates, especially if you lean on cached prompts.
  3. Plan a side-by-side test with your current model, such as GPT-6.1 Terra or Claude models, on your own tasks.
  4. If you do security work, expect stricter refusals at first and check whether you qualify for Daybreak.

Everyday users of ChatGPT may mostly notice better slides and documents. You can also follow our latest AI news for the confirmed launch details once OpenAI makes them official.

Sources

Primary source: OpenAI’s Introducing GPT-6.2 Astra and Sol page. Additional context from OpenAI’s Astra for Law update.

Conclusion

OpenAI’s GPT-6.2 Astra and Sol page is already live on its website, even though the company has not formally announced the models. It promises a stronger frontier model, a cheaper everyday sibling and lower Astra prices than GPT-6 Astra. It also flags real safety trade-offs, including tighter cyber limits and a new self-improvement rating. Until OpenAI confirms the release, treat it as a detailed preview rather than a finished launch.

Newsletter

New AI tools in your inbox

The best new tools and the AI news that matters, in one short email. Unsubscribe any time.

Free · No spam · Unsubscribe any time

✓ Looks like you’re subscribed

Loading Google’s sign-up form…

Looks like you’re already subscribed

Thanks for reading! New AI tools and the AI news that matters will land in your inbox.

Not subscribed yet? Refresh the page to load the sign-up form.

The sign-up form couldn’t load. If you use an ad or content blocker, allow news.google.com and reload the page.

Free instant check before you pay. Takes a few seconds.

Scroll to Top