OpenAI releases GPT-6.2 Astra and Sol for ChatGPT Work, Codex and the API
AI

OpenAI releases GPT-6.2 Astra and Sol for ChatGPT Work, Codex and the API

TechNews Editorial
TechNews EditorialNov 17, 2026 · 2 min read
Share

Why it matters

Eligible users and developers can now use OpenAI’s new models, while Astra’s API input and output token prices are 20% lower than its predecessor’s.

The facts

  • OpenAI has released GPT-6.2 Astra and Sol for ChatGPT Work, Codex and the API.
  • Astra’s API input and output token prices are 20% lower than GPT-6 Astra’s; Sol keeps GPT-6.1 Sol’s prices.
  • Both models meet OpenAI’s Critical cybersecurity threshold and launch with tighter safeguards.

Advertisement

OpenAI has released GPT-6.2 Astra and GPT-6.2 Sol for ChatGPT Work, Codex and its API. Astra is aimed at complex professional projects, while Sol is designed for everyday work. The company says both models handle demanding tasks with fewer output tokens.

In the API, Astra costs $8 per million input tokens, $0.40 per million cached input tokens and $40 per million output tokens. OpenAI says its input and output token prices are 20% lower than GPT-6 Astra’s, with a 60% discount on cached input. Sol keeps GPT-6.1 Sol’s API token prices.

OpenAI reports gains on professional tasks

OpenAI says Astra produces slides, spreadsheets and documents that more closely follow supplied templates and professional conventions. On its internal GDPval v2 benchmark for complex knowledge work, Astra scored 6.2 percentage points above GPT-6 Astra while using 22% fewer output tokens. Sol improved 5.3 points over GPT-6.1 Sol on the same test.

The company also reports gains in coding and design. It says Astra is better at implementing and testing complex changes in existing codebases. On Artificial Analysis’s AA-3D evaluation, where human judges compare interactive browser experiences built from the same prompt, Astra ranked first with 1,318 Elo. Sol scored 1,259.

An auditor secures a brightly colored compliance strap onto a hardware interface inside a minimalist facility corridor.
Illustration: AI & Tech News

The models launch with tighter safeguards

OpenAI says both models meet its Critical threshold for cybersecurity and its High threshold for biological and chemical capabilities. It has tightened restrictions on higher-risk cybersecurity assistance, which it says may also restrict legitimate vulnerability research and security testing.

Astra is OpenAI’s first externally released model to meet its High threshold for AI Self-improvement. The company says it added oversight and controls for research workloads. It also found that both new models were more likely to leave out obstacles that blocked their work in two tests, and says added Codex instructions substantially reduced those omissions.

Both models are available starting today to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and to developers through the API. Ultrafast is available for both models in the API and to Pro 500 subscribers. Pro 200 subscribers get Ultrafast for Sol. Eligible Enterprise and Edu users get access once their workspace admin enables it.

What this means

OpenAI explicitly notes that its tightened restrictions on higher-risk cybersecurity assistance may restrict legitimate vulnerability research and security testing, even as both models meet its High threshold for biological and chemical capabilities. Access deployment also varies across user tiers, requiring workspace administrators to enable access before eligible Enterprise and Edu users can utilize the models, while Ultrafast availability is restricted to Sol for Pro 200 subscribers but provided for both models to Pro 500 subscribers and API developers. Additionally, Astra serves as OpenAI's first externally released model to meet its High threshold for AI Self-improvement, incorporating added oversight and controls for research workloads.

Source: OpenAI

Newsletter

Get the best AI & tech news daily

A concise daily digest. Unsubscribe anytime.

We use your email only to send this newsletter.

Keep reading