Skip to content
Advertisement
OpenAI launches GPT-6.2 Astra and Sol models for ChatGPT and API

OpenAI launches GPT-6.2 Astra and Sol models for ChatGPT and API

OpenAI splits its latest release into a heavy-duty model for complex work and a cheaper one for everyday tasks

OpenAI has released two new models at once. GPT-6.2 Astra is built for demanding work, and GPT-6.2 Sol handles everyday tasks.

Both arrived on November 17, 2026, and both are live in ChatGPT and the API.

Two models, two jobs

Astra is aimed at high-complexity work like scientific research, coding and large, multi-step projects.

Sol is tuned for efficiency on common tasks and general workflows, where speed and cost matter more than raw horsepower.

Both models promise stronger inference. That means they reason through a problem more capably before answering.

OpenAI also says both aim to reach better results while using fewer output tokens. Tokens are the small chunks of text a model produces, and they are how API usage gets billed. Fewer tokens for the same answer means a smaller invoice.

Advertisement

Who gets access, and how

Inside ChatGPT, the models are rolling out to Plus, Pro, Enterprise and Edu users. They appear in ChatGPT Work and in Codex, OpenAI’s coding product.

Developers can call them through the API using the identifiers gpt-6.2-astra and gpt-6.2-sol.

Pro 500 subscribers receive Ultrafast access. Pro 200 subscribers get it first for GPT-6.2 Sol.

How the models stack up

Independent evaluator Artificial Analysis ranks GPT-6.2 Astra second on its Intelligence Index. The only model ahead of it is Anthropic’s Claude Fable 5.2.

Astra scored especially well in categories covering 3D design, science, knowledge and robotics. Specific benchmarks where it performed strongly include AA-3D and Terminal-Bench-Science 0.1.

On cost-per-task and output-token efficiency, the new models share or lead the frontier compared with competitors.

Pricing keeps the familiar hierarchy. Astra remains the premium model, while Sol costs about one-fifth as much per token.

Tighter guardrails for sharper tools

OpenAI has acknowledged more capable models bring more capable misuse by adding stricter safeguards in higher-risk areas.

Cybersecurity is the standout. The company cited the models’ stronger capabilities in that domain as the reason for tighter controls.

A fast-moving release cycle

OpenAI debuted its flagship GPT-6 Astra on September 3, 2026, and has shipped a series of updates since.

Roughly two and a half months separate the GPT-6 Astra debut from this 6.2 pairing.

Disclosure: This article was edited by Diego Almada Lopez. For more information on how we create and review content, see our Editorial Policy.