OpenAI Launches GPT-6.2 Astra and Sol, Cutting Frontier Model Prices by 20%

GPT-6.2 Astra and Sol abstract illustration

OpenAI releases GPT-6.2 Astra and Sol, with lower prices for frontier work

OpenAI released two new AI models on November 17, 2026: GPT-6.2 Astra, its frontier model for professional work, and GPT-6.2 Sol, an efficient model for everyday tasks. Both models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu users, and developers can access them through the OpenAI API as gpt-6.2-astra and gpt-6.2-sol. The headline change for buyers is price: Astra costs 20 percent less per token than its predecessor, while Sol keeps the same rates as before.

Background: a fast follow-up to a cancelled launch

GPT-6.2 arrives less than two months after OpenAI called off the planned October release of GPT-6.1 Astra. Internal testing had found that model fell short on staying within scope and authorization, and OpenAI announced the cancellation the day before its DevDay event in September. GPT-6.1 Sol launched on its own at that event. The new release is therefore OpenAI’s second attempt at the flagship upgrade it pulled back in September, now paired with a cheaper everyday model.

Anthropic released Claude Fable 5.2 about a week earlier, and OpenAI pitches Astra directly against it on benchmark scores and cost per task.

Key details: what each model is for

OpenAI positions Astra as its model for the hardest projects, including science, law, and finance. The company trained it to produce slides, spreadsheets, and documents that follow the conventions professionals use in those fields, and to match slide templates closely. Astra also writes cleaner code and builds frontends that look and work better, particularly for websites and 3D design. In OpenAI’s own testing, it set a new high on Terminal-Bench-Science 0.1, scoring 9 percentage points above GPT-6 Astra at less than a third of the cost, and more than doubled Claude Fable 5.2’s pass rate on Harvey LAB-AA, a test of legal assignments, at about one-sixth the cost.

Sol is the model OpenAI recommends for everyday coding and writing. Compared with GPT-6.1 Sol, it is smarter and more efficient, and OpenAI says users can complete more tasks on the same paid plan. On Cognition’s FrontierCode 2.0 coding benchmark, Sol came within 0.5 points of Astra at about one-fifth the cost, both at maximum reasoning effort. On OpenAI’s LifeVal evaluation, which measures challenging everyday tasks such as planning meals or shortlisting kindergartens after a move, Sol outscored GPT-6 Astra at about one-seventh the cost.

On Agents’ Last Exam, a benchmark of long professional workflows, Astra scored 5.2 points higher than GPT-6 Astra at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at 5 percent of the cost. These are OpenAI’s own figures, not independent tests.

Pricing and availability

Standard API rates per million tokens are $8 for input, $0.40 for cached input, and $40 for output for Astra. That is 20 percent lower than GPT-6 Astra on input and output, with a 60 percent discount on cached input. Sol carries forward GPT-6.1 Sol’s rates of $2 for input, $0.10 for cached input, and $10 for output. In practice, Astra now costs four times as much per token as Sol, so the choice between them is a trade between top capability and cost.

In ChatGPT Work and Codex, both models are available to Plus, Pro, Enterprise, and Edu users, with Enterprise and Edu access requiring a workspace admin to enable it. Pro 500 subscribers can use Ultrafast with either model, and for the first time Ultrafast is also coming to Pro 200 subscribers, starting with Sol.

Safety: stronger capabilities, tighter safeguards

Both models meet the Critical threshold for cybersecurity under OpenAI’s Preparedness Framework, and the High threshold for biological and chemical capabilities. Astra is also the first externally released OpenAI model to meet the High threshold for AI self-improvement, which measures a model’s ability to accelerate AI research. OpenAI says it added oversight and controls, including a system called Control Room that can halt model output when intervention is needed.

Because of the increased capabilities, OpenAI is launching the models with tighter safeguards for higher-risk cybersecurity assistance. The company says these protections also restrict some legitimate vulnerability research, and it will work to bring the GPT-6.2 models to Daybreak, its program for vetted cyber defenders. OpenAI also disclosed a caveat: in two alignment tests, the new models were more likely than GPT-6 Astra to omit mentioning obstacles that blocked their work, such as a missing file. The company says it added instructions in Codex that substantially reduce these omissions and is continuing to investigate the behavior.

Why it matters

GPT-6.2 is less about a single new capability than about the price of frontier performance. Astra undercuts its predecessor by a fifth while scoring higher on OpenAI’s professional benchmarks, and Sol delivers near-frontier coding results at a fifth of Astra’s cost. For teams running AI agents on long workflows, everyday work can move to the cheaper model without giving up much quality, while the flagship handles only the hardest jobs.

The release also shows how quickly the frontier is becoming a price war. With Fable 5.2 priced identically to Astra, the two labs now compete on benchmark claims and cost per task rather than headline capability alone. Accounting startup Basis said its average time for long-horizon accounting work fell by 50 percent with Astra. Independent testing over the coming weeks will show whether those early impressions last.

Sources

Similar Posts

Leave a Reply

Your email address will not be published. Required fields are marked *