OpenAI began rolling out two new models, GPT-6.2 Astra and GPT-6.2 Sol, on November 17, 2026 (US time). Both are available starting the same day in ChatGPT Work, the coding assistant Codex, and the developer API. The company recommends Astra for the hardest projects and Sol for everyday coding and writing.
Astra Targets Demanding Work, Sol Handles Everyday Tasks
According to OpenAI's GPT-6.2 overview page, GPT-6.2 Astra (Astra) is the stronger choice for complex reasoning. OpenAI says it can work through a research question across multiple sources and check its conclusions against the data. In Codex, the company's example is tracing a bug across services, fixing it, and testing the affected workflows.
Another example OpenAI gives for Astra is building a slide deck. Users provide a report and their slide template, then ask the model to present the findings in their team's layout and writing style.
GPT-6.2 Sol (Sol), meanwhile, handles everyday coding, writing, and work that takes several steps. OpenAI says that compared with GPT-6.1 Sol, it lets users complete more tasks on the same paid plan. Its Codex examples include reviewing a pull request and revising code in response to feedback.
In ChatGPT, the Models Live in Work
To use the two models in ChatGPT, users open Work. They are available on the Plus, Pro, Business, Enterprise, and Edu plans, on the web, desktop, and mobile. On Enterprise and Edu, admins must first enable the models in workspace settings.
On supported consumer plans, new Work chats default to Sol at the "Light" reasoning effort. Enterprise workspaces with model access default to Sol at "Medium." Users can pick the model and reasoning effort themselves under "Advanced." Codex is available in both the app and the CLI (command-line tool), and it shares usage with Work.
API Pricing: Astra 20% Cheaper, Sol Unchanged
In the API, the models are called "gpt-6.2-astra" and "gpt-6.2-sol." Standard pricing listed on OpenAI's API pricing page, per 1 million tokens for prompts up to 272,000 input tokens, is as follows.
| Model | Input | Cached input | Output |
|---|---|---|---|
| GPT-6.2 Astra | $8.00 | $0.40 | $40.00 |
| GPT-6 Astra | $10.00 | $1.00 | $50.00 |
| GPT-6.2 Sol | $2.00 | $0.10 | $10.00 |
| GPT-6.1 Sol | $2.00 | $0.10 | $10.00 |
Astra's input and output prices are 20% lower than GPT-6 Astra's, while Sol keeps the same pricing as GPT-6.1 Sol. For Batch processing, Astra costs $4.00 input / $20.00 output and Sol costs $1.00 input / $5.00 output.
OpenAI's API changelog states that both models support the Responses API and Chat Completions, although tool calling requires the Responses API. On the same day, OpenAI also opened a public beta of background processing with a two-hour or eight-hour completion window. Token rates are 30% below Standard for the two-hour window and 40% below for the eight-hour window.
Ultrafast Runs Up to 4.5x Faster, at a Premium
Both GPT-6.2 models support the high-speed "Ultrafast" mode at launch. OpenAI says Ultrafast generates tokens at up to 4.5 times the speed of Standard on the same model. In ChatGPT, the $200-per-month Pro plan gets Ultrafast with Sol only, while the $500-per-month Pro plan gets it with either model. Eligible Enterprise and Edu workspaces can use both once an owner enables it.
In the API, Ultrafast costs 3 times the Standard rate through February 28, 2027, and 6 times the Standard rate starting March 1, 2027 (Pacific time). Ultrafast also supports US data residency only. Standard, Fast, and Batch processing support both US and EU residency.
Developers Should Check Their Settings When Migrating
OpenAI's migration guide lists several changes that come with GPT-6.2. Parameters such as temperature are not supported and must be omitted. For new integrations, OpenAI recommends starting with "medium" reasoning effort. In addition, if a developer edits the beginning of a prompt or switches to a lower-tier model mid-conversation, the API stops carrying over the model's intermediate reasoning. OpenAI says answer quality can decline and token use can rise in that case.
Our Take
The Astra price cut looks like a move to make OpenAI's top model easier to build into everyday development. Cached input now costs four-tenths of the previous price, so workloads that resend the same prompt prefix should benefit the most.
Speed, however, comes at a steep cost. From March 2027, Ultrafast will cost 6 times Standard, which puts Astra's output at $240 per 1 million tokens. That said, work that is not urgent can be routed to Batch or background processing at lower rates. Developers now have more room than before to match models and processing modes to each task.
Summary
OpenAI released two GPT-6.2 models at once, the top-tier Astra and the everyday Sol, across ChatGPT Work, Codex, and the API. The update matters both to paid ChatGPT subscribers and to developers using the API. OpenAI says details on benchmarks and safety evaluations are available in its release announcement.