AI models

GPT-6.2 Astra and Sol: what changes with OpenAI's new flagship models

On Tuesday, November 17, 2026 OpenAI released GPT-6.2 Astra and GPT-6.2 Sol, the two most capable models in its lineup. Astra is built for the most demanding projects, Sol for everyday work at a lower cost. Both are already in ChatGPT Work and Codex for subscribers and in the API, with Astra now 20% cheaper. Here is what changes, what it costs, who can use them and which numbers are only claimed.

Published on 7 min readBy the editorial team of The AI Creator System

Leggi questo articolo in italiano

AI-generated cover: two glowing cyan and violet spheres, one large and one smaller, on a dark blue background, title GPT-6.2 Astra and Sol
Cover image generated with artificial intelligence. How we use AI

What GPT-6.2 Astra and GPT-6.2 Sol are

GPT-6.2 Astra and GPT-6.2 Sol are OpenAI's two new top-tier models, released on Tuesday, November 17, 2026: Astra is the flagship for the hardest work, Sol is the cheaper option for everyday tasks. OpenAI announced them on its official developer forum, and the launch was confirmed by several outlets.

To find your way around OpenAI's lineup, remember its four model classes, from lightest to most powerful: Luna, Terra (updated in late October with GPT-6.1 Terra), Sol and Astra. With this release the two top classes move to version 6.2.

  • GPT-6.2 Astra replaces GPT-6 Astra, released in September. According to The Decoder, it targets demanding professional work in science, law and finance.
  • GPT-6.2 Sol replaces GPT-6.1 Sol and is the more affordable option for everyday coding and writing.

OpenAI stresses one practical point: both models finish demanding tasks with fewer output tokens, which the company says helps subscribers stay within their usage limits.

What changes in ChatGPT and Codex

GPT-6.2 Astra remains OpenAI's recommended model for the hardest projects, while GPT-6.2 Sol covers most everyday work for less. Both have been available since November 17, 2026 in ChatGPT Work and in Codex, OpenAI's coding environment.

Among the changes reported by 9to5Mac:

  • slides from a template: Astra can use a screenshot of a slide as a template, matching its layout, colors and font size;
  • cleaner code in Codex: Astra is better at making and testing changes in existing projects; early testers report a tidier code style, more focused tests and fewer rounds of revision;
  • Sol close to Astra on code: on FrontierCode 2.0, which checks whether changes to real codebases are good enough to merge, Sol comes within half a point of Astra at max effort, for about a fifth of the cost (a claimed figure).

Ultrafast mode generates text up to 4.5 times faster than standard but uses more of your allowance; you switch to it from the model picker. Subscribers on the $500-a-month Pro plan get it on both models, while those on the $200 Pro plan get it for the first time, starting with Sol. Eligible Enterprise and Edu workspaces get it once an admin turns it on.

Pricing and availability: who can use them and what they cost

GPT-6.2 Astra and GPT-6.2 Sol have been available since November 17, 2026 to Plus, Pro, Enterprise and Edu subscribers in ChatGPT Work and Codex, and to developers in the API as gpt-6.2-astra and gpt-6.2-sol. The sources we checked do not mention ChatGPT's free plan.

API price (per million tokens)GPT-6.2 AstraGPT-6.2 Sol
Input$8$2
Cached input$0.40$0.10
Output$40$10

For Astra that is a 20% cut on GPT-6 Astra's standard prices; Sol keeps GPT-6.1 Sol's pricing. Input and output prices are in OpenAI's announcement, cached prices are reported by The Decoder.

If you use ChatGPT on a subscription, per-token pricing does not affect you directly. It matters to anyone building automations, chatbots and agents on the API: there, a model that does the same job with fewer tokens, at a lower price, changes the cost of every conversation.

The numbers: what OpenAI claims and what independent testing says

OpenAI claims clear gains on professional work at a lower cost per task; independent rankings put GPT-6.2 Astra among the top models, but not first everywhere. Here are the figures claimed by OpenAI, as reported by the sources:

  • GDPval-v2 (long professional tasks): Astra gains 6.2 percentage points over its predecessor while using 22% fewer output tokens;
  • Agents' Last Exam: Astra scores 5.2 points above GPT-6 Astra, at a much lower cost per task;
  • FrontierCode 2.0: Sol within half a point of Astra at about a fifth of the cost.

External measurements, reported by The Decoder, are more nuanced:

  • Artificial Analysis Intelligence Index: Astra 62, second behind Claude Fable 5.2 (65); Sol 57. Gemini 4 Carbon scores 60 and Claude Opus 5.5 59;
  • Epoch Capabilities Index: Astra 174, tied with Claude Fable 5.2;
  • Vals Index: Astra seventh, Sol ninth;
  • hallucinations: on the AA-Omniscience test at max effort Astra is wrong 59% of the time and Sol 55%. Artificial Analysis warns these are unusually hard questions, not an everyday error rate.

On safety, OpenAI rates both models at the Critical level for cybersecurity under its Preparedness Framework, as it did GPT-6 Astra, and Astra is the first externally released model at the High level for AI self-improvement. As a result, help with higher-risk cybersecurity tasks is more restricted by default; less restrictive access will go through the Daybreak program for approved defenders.

In short: more efficient and cheaper, with Sol closing in on Astra for code; in independent rankings the race with Claude and Gemini is still open.

What it means for you if you create content or work for yourself

If you work alone or in a small team, GPT-6.2 mostly means two things: long jobs finished while using less of your allowance, and the option to use Sol for almost everything, keeping Astra for the hard cases.

  1. Start with Sol. For emails, copy, small scripts and reviews it is almost always enough; switch to Astra when the task is long, technical or draws on many sources.
  2. Use Astra for documents with a set format. Giving it a slide screenshot as a template is the quickest way to get presentations that match your style.
  3. Skip Ultrafast when you do not need it. It is handy when you are waiting on the answer, but it burns more allowance: keep it for live work.
  4. Check the facts. Error rates on hard questions are still high: verify numbers, names and quotes before you publish.

For ready-made prompts and procedures to give ChatGPT in your everyday work, there is the AI Business Toolkit; if you handle customer messages and requests, Customer Care AI shows you how to automate them, and for quotes and sales there is AI Sales Machine. If you also work with Claude, which still leads the Artificial Analysis index, Claude Business OS covers emails, meetings, documents and numbers.

To compare the alternatives, also read about Google's Gemini 4 Hydrogen and Claude Meetings, plus our ChatGPT Images guide. For a complete method to create with AI there is The AI Creator System, and here is how we use artificial intelligence.

Frequently asked questions

What is GPT-6.2 Astra?

It is OpenAI's flagship model, released on November 17, 2026 to replace GPT-6 Astra. It is built for the most demanding professional projects, for example in science, law and finance, and for complex coding in Codex.

What is the difference between GPT-6.2 Astra and GPT-6.2 Sol?

Astra is the most capable and the most expensive; Sol costs about a quarter as much in the API and is built for everyday coding and writing. On a real-code benchmark (FrontierCode 2.0) OpenAI claims Sol comes within half a point of Astra at about a fifth of the cost.

How much does GPT-6.2 cost in the API?

GPT-6.2 Astra costs $8 per million input tokens and $40 per million output tokens; GPT-6.2 Sol costs $2 and $10. For Astra that is 20% less than GPT-6 Astra.

Is GPT-6.2 available on free ChatGPT?

The launch sources from November 17, 2026 mention Plus, Pro, Enterprise and Edu subscribers in ChatGPT Work and Codex, plus the API. The free plan is not mentioned.

What is ChatGPT's Ultrafast mode?

It is a mode that generates text up to 4.5 times faster but uses more of your allowance. With GPT-6.2 it is available on both models on the $500 Pro plan and, for the first time, on the $200 Pro plan starting with Sol.

Read next

Sources checked on November 18, 2026