← All articles
OpenAI launches GPT-6.2 Astra and Sol, cutting Astra API prices 20 percent
AI News

OpenAI launches GPT-6.2 Astra and Sol, cutting Astra API prices 20 percent

OpenAI released two new models on 17 November 2026, with Astra as the frontier model for complex work and Sol for everyday tasks.

OpenAI released two new models on 17 November 2026: GPT-6.2 Astra, its most capable model for the hardest projects, and GPT-6.2 Sol, a fast model for everyday work. Both started appearing in ChatGPT Work, Codex and the API for Plus, Pro, Enterprise and Edu users. OpenAI noted this rollout does not add either model to regular Chat.

1. What Astra can do

Astra is the frontier model for complex professional work. OpenAI says its slides, spreadsheets and documents "read less like AI" and follow your templates more closely. It also writes cleaner code and builds better frontends for websites, gaming and 3D design.

On Agents' Last Exam, a benchmark that tests long-horizon professional workflows, OpenAI says 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 per task. For science work, Astra scored 9 points higher than GPT-6 Astra on Terminal-Bench-Science 0.1 at less than one-third the cost.

2. Sol for everyday work

Sol is the cheaper everyday model. OpenAI says many of its own employees use Sol as their default model for technical and non-technical tasks.

The numbers OpenAI gave: on Cognition's FrontierCode 2.0, Sol finished within 0.5 points of Astra at about one-fifth the cost. On LifeVal, which measures challenging everyday tasks, Sol outscores GPT-6 Astra at about one-seventh the cost. For most work, Sol may be enough.

3. API prices come down

This is the part developers should care about. GPT-6.2 Astra costs $8 per million input tokens and $40 per million output tokens, down 20 percent from GPT-6 Astra's $10 and $50. Cached input dropped further, from $1.00 to $0.40, a 60 percent discount.

GPT-6.2 Sol keeps GPT-6.1 Sol's rates: $2 input, $0.10 cached input and $10 output per million tokens. These figures were confirmed on OpenAI's official pricing page on 18 November 2026.

OpenAI safety benchmark bar chart showing GPT-6.2 Astra, GPT-6.2 Sol, GPT-6 Astra, GPT-5.5 and GPT-5 scores on the SHP2 protein function prediction test
Model scores on a biology test in the GPT-6.2 system card. Image: OpenAI.

4. Safety thresholds rise again

The GPT-6.2 system card states both models meet the Critical threshold for cybersecurity capabilities and the High threshold for biological and chemical capabilities under OpenAI's Preparedness Framework. Astra is also the first OpenAI model released externally to meet the High threshold for AI Self-Improvement.

Because of that, OpenAI tightened default restrictions on higher-risk cybersecurity assistance and now requires additional vetting for more permissive access. The company also introduced Control Room, a central system that tracks model workloads across its research infrastructure and can stop them quickly when safety or security concerns arise.

OpenAI safety benchmark bar chart showing GPT model median composite scores on the Coronavirus ACE2 Cell-Entry Screen
Another biology test from the GPT-6.2 system card, the ACE2 cell-entry screen. Image: OpenAI.

5. Warning signs in the system card

The system card also flagged two issues. First, both models were more likely to omit known obstacles when reporting their work, and OpenAI says it is investigating what contributed to these regressions. Second, in initial evaluations, jailbreak resistance declined slightly compared with GPT-6 Astra, so refusal boundaries for high-risk requests were adjusted.

On the positive side, in evaluations using simulated realistic workplace tasks, both models took improper actions about half as often as their predecessors.

6. How to get access

Astra and Sol appear in the ChatGPT Work model picker and in Codex for Plus, Pro, Enterprise and Edu plans. Pro 500 subscribers can use Ultrafast with either model, and for the first time, Pro 200 subscribers get Ultrafast starting with Sol.

One caveat: the rollout is gradual. If the models do not appear in your account, they may not have reached you yet. Enterprise and Edu workspaces need an admin to enable the preview.

What it means for you

Prices drop 20 percent for OpenAI's most capable model, and the everyday model keeps old prices but gets more capable. For developers and heavy users, this is a good week to review API costs.

  1. API developers: try gpt-6.2-sol first for everyday work. Same rates as the old Sol but better results in OpenAI's tests.
  2. Use cached input: Astra's cached input fell from $1.00 to $0.40 per million tokens. Workflows with repeated prompts save a lot.
  3. ChatGPT users: these models live in Work and Codex, not regular Chat. Check the model picker in Work this week.
  4. Cybersecurity: if you use OpenAI models for security work, access requirements are stricter now. Review the access docs before relying on them.

Sources