Loading market data...

OpenAI Releases GPT-6.2 Astra and Sol, Splitting High-Complexity and Everyday Work

OpenAI Releases GPT-6.2 Astra and Sol, Splitting High-Complexity and Everyday Work

OpenAI has released two new models, GPT-6.2 Astra and GPT-6.2 Sol, making both available through ChatGPT and the API. Astra is built for high-complexity tasks; Sol is aimed at everyday work and cost-efficient solutions. The company says the pair brings improvements in reasoning, coding, scientific research, and professional workflows, and that processing costs are lower.

Two models, two jobs

The split is deliberate. Astra takes the harder work — the kind of tasks that have typically pushed models toward longer reasoning chains and heavier compute. Sol covers the routine, high-volume requests where cost per call matters as much as output quality. OpenAI describes the release as balancing high-complexity tasks with cost-efficient solutions across industries, though it hasn't detailed which sectors it expects to adopt which model first.

Both models are live for ChatGPT users and API customers now. That dual availability matters: API customers can wire Astra or Sol into their own products, while ChatGPT users get the improvements without touching a line of code.

What OpenAI says improved

The company lists reasoning, coding, scientific research, and professional workflows as the areas where the new models perform better. It also says GPT-6.2 Astra performs better on certain benchmarks, but the announcement excerpt didn't include the specific numbers or the benchmarks themselves. Until those figures are published, the performance claims rest on OpenAI's own description rather than independent testing.

The cost reduction is the more concrete claim. Lower processing costs, if they hold at scale, change the math for teams running large volumes of requests — especially those who previously reserved expensive models for tasks that Sol can now handle.

The pricing question

OpenAI hasn't disclosed per-token pricing for either model in the release materials reviewed. That leaves API customers without the number they need most when deciding whether to migrate existing workloads from older models. The company's framing — Astra for complexity, Sol for cost efficiency — suggests Sol will sit at a lower price point, but the actual rates haven't been confirmed.

For ChatGPT users, the change is less about price and more about which model answers a given prompt. OpenAI hasn't said how the two models are routed inside ChatGPT, or whether users will be able to choose between them.

What happens next

The immediate next step is watching for benchmark data and pricing details from OpenAI, and for independent evaluations of Astra's reasoning and coding performance. Developers weighing a switch will want both before committing production traffic to either model. The models are available now, so early adopters can test them — but the case for a full migration still hinges on numbers OpenAI hasn't yet put on the record.