Model: GPT-6.2 Astra & GPT-6.2 Sol
On 17 November 2026, OpenAI released GPT-6.2 Astra and GPT-6.2 Sol. GPT-6.2 Astra is a frontier reasoning model that handles complex research and professional work, while GPT-6.2 Sol is a faster, lower-cost model for multi-step everyday tasks. Both arrive in ChatGPT Work, Codex and the API — the places where buyers now research vendors.
Source: OpenAI — model release notes and ChatGPT release notes (17 November 2026)
- On 17 November 2026, OpenAI released GPT-6.2 Astra (frontier reasoning) and GPT-6.2 Sol (speed and lower cost).
- Both are in preview in ChatGPT Work and Codex desktop for Plus and Pro, and in the API as
gpt-6.2-astraandgpt-6.2-sol. - Regular Chat keeps its existing model choices for now — the change lands first in professional research and in apps built on the API.
- OpenAI says both models do better on demanding tasks while generating fewer output tokens, so answers may get tighter and shortlists shorter.
- A model change is a re-baseline moment: check how your brand shows up in Work-style research before and after.
What OpenAI actually shipped
On 17 November 2026, OpenAI introduced two new models. In OpenAI’s words, Astra is its “frontier model for complex work across research, software development, and other professional fields”, while Sol “combines strong reasoning with speed and efficiency for a wide range of everyday tasks.”
They update the pair we covered earlier this autumn — GPT-6 Astra in September and GPT-6.1 Sol in October — and arrive three weeks after GPT-6.1 Terra. OpenAI’s model line is now updating roughly every few weeks.
Where can you use GPT-6.2 today?
Both models are in preview in ChatGPT Work and Codex desktop. On Plus or Pro you pick either one in the Work model picker or the Codex model selector; in a managed workspace, what you see depends on your seat and workspace settings.
Developers get them in the OpenAI API through the Responses and Chat Completions endpoints. Both accept text and images and return text. One detail matters for brands: OpenAI notes that “Regular Chat keeps its existing model choices.” The everyday consumer chat window is not switching yet.
Why does Work-mode research matter?
Work is where professionals run long research tasks: comparing suppliers, drafting shortlists, summarising a market. When Astra runs that research, it decides which companies make the shortlist and which sources back it up. For B2B brands, that is often the most valuable answer of all.
“Compare the top three payroll providers for a 200-person UK firm” or “Which logistics platforms integrate with Shopify?” are exactly the prompts a frontier research model handles. A stronger model can read more pages, weigh more evidence and be more selective about who it names — the pattern we saw with deep research in Work and Codex.
What fewer output tokens could change
OpenAI says both models improve on their predecessors’ results on demanding tasks “while generating fewer output tokens.” In plain terms, they say more with less. That is good for users, but it can squeeze the space in an answer where brands get named.
A tighter answer tends to mean a shorter list. If a model used to name six options and now names three, the brands in fourth to sixth place fall out entirely. Being clearly the best fit for a specific question becomes more important than being broadly relevant. We are not claiming this will happen for every prompt — it is a shift worth measuring, not assuming.
GPT-6.2 Astra vs GPT-6.2 Sol
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on 17 November 2026.
Why the API launch matters too
The API release puts the same models inside other companies’ products: comparison sites, shopping assistants, support bots and internal research tools. Sol’s lower cost makes it an easy default for apps that answer a lot of questions, so its view of your category can spread well beyond ChatGPT itself.
That is why a model release matters even when the consumer chat window stays the same. Your brand can be described by GPT-6.2 in a dozen places you never see. The facts it relies on — your site, reviews, listings and the sources it cites — are the same either way.
What GPT-6.2 does not change
It is one vendor. GPT-6.2 tells you nothing about how Gemini 4 Carbon, Claude, Perplexity, Copilot or Google AI Overview answer the same questions, and those models often name different brands for the same prompt.
It also does not rewrite the rules for being chosen. Models still favour brands that match the question clearly and are backed up consistently across the web — the logic we explain in how LLMs decide which brands to cite. A better model reads that evidence more carefully; it does not invent evidence you have not published.
What should brands do now?
Treat GPT-6.2 as a prompt to tighten your evidence, not a reason to panic. Make sure the pages a research model would read state plainly what you do, who you serve, what it costs and how you compare. Keep those facts consistent across third-party listings, and make sure AI crawlers can actually reach your site.
Then measure. Today is 21 November 2026, and a short checklist will show where you stand:
- List your research prompts. Write down the comparison and shortlist questions a buyer in your category would ask.
- Run them in Work. If you have Plus or Pro, try them with GPT-6.2 Astra and with an older model, and note which brands each names.
- Count the shortlist. Check whether answers name fewer options than before, and where you land.
- Check the sources. See which pages the model leans on and whether your own content is among them.
- Read the description. Look at how it describes you, not just whether it names you.
- Compare across models. Repeat in Gemini, Claude, Perplexity and Copilot, since each picks differently.
- Re-run in a month. One test is a snapshot; the trend is what tells you if anything shifted.
How reconnAI tracks new models
reconnAI monitors how the leading AI models — ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — answer questions across regions, and re-baselines that tracking when a platform ships a model that changes behaviour. A new frontier pair from OpenAI is exactly that kind of moment.
It follows a busy few weeks, from GPT-6 Rosalind and Gemini 4 Hydrogen to ChatGPT Wallet. For the bigger picture, see building a GEO strategy and is your brand ready for AI search? If you want to see how these models represent your brand, talk to our team, explore our AI answer tracking, or read our guide to ChatGPT ads for where paid placement is heading.
GPT-6.2: your questions answered
What did OpenAI release? On 17 November 2026, OpenAI released GPT-6.2 Astra, a frontier model for complex professional work, and GPT-6.2 Sol, a faster, lower-cost model for multi-step everyday tasks.
Is GPT-6.2 in regular ChatGPT? Not yet. Both models are in preview in ChatGPT Work and Codex desktop for Plus and Pro users, and in the OpenAI API. Regular Chat keeps its existing model choices.
What is the difference between Astra and Sol? Astra is built for advanced reasoning, coding and demanding professional work. Sol trades some depth for speed and lower cost while still handling tasks that take several steps.
Will GPT-6.2 change which brands ChatGPT recommends? It can. A new model reads and weighs sources differently, and OpenAI says both generate fewer output tokens, which may mean shorter shortlists. The only way to know for your brand is to test before and after.
How do I track my visibility across models? Baseline the questions your customers ask, run them across ChatGPT, Gemini, Claude, Perplexity and Copilot, and watch how answers change over time — which is what reconnAI automates.
About reconnAI
reconnAI tracks how the major AI models represent topics and sources across ChatGPT, Claude, Gemini, Perplexity, Copilot and Google AI Overview — across multiple regions. We monitor how those models answer and how they change over time, so you can stay ahead of shifts in the AI landscape.
Last updated: 21 November 2026.