OpenAI shipped two new models on Tuesday, November 17 — and the headline is not just that they are smarter. GPT-6.2 Astra, the new frontier model for demanding professional and scientific work, and GPT-6.2 Sol, its cheaper everyday sibling, are live now in ChatGPT Work and Codex for Plus, Pro, Enterprise and Edu users, and in the API as gpt-6.2-astra and gpt-6.2-sol. One important catch: this rollout does not add either model to regular Chat.

Two glowing AI cores — a large white-gold flagship core and a smaller fast teal core — over a dark futuristic workspace, illustrating OpenAI's GPT-6.2 Astra and Sol launch

OpenAI says the pair is built to do more inside the same usage limits: “Both models deliver better results on demanding tasks with fewer output tokens, so you can get more done within your usage limits.” Astra is described in OpenAI’s announcement as “a frontier model built for your hardest projects,” setting new highs on professional-work and scientific-research evaluations at less than a third of the cost per task of GPT-6 Astra.

The numbers: Astra at 62, Sol close behind at 57

Independent evaluator Artificial Analysis puts GPT-6.2 Astra at 62 on its Intelligence Index, up from 57 for GPT-6 Astra — at a lower weighted cost per task, $4.00 versus $6.24. Sol jumps to 57, up from 55 for GPT-6.1 Sol, at just $1.01 per task. On Cognition’s FrontierCode 2.0 benchmark of merge-ready changes to real codebases, OpenAI says Sol nearly matches Astra at about one-fifth of the cost. Reports on the system card also say Astra is the first model OpenAI has released that it rates High for AI self-improvement, and both models launch with tighter safeguards around higher-risk cybersecurity help.

Pricing: the flagship got 20% cheaper

Astra’s standard API rates fall 20% from its predecessor, to $8 per million input tokens and $40 per million output tokens, with cached input at $0.40. Sol holds its previous pricing at $2 input and $10 output. Both accept text and images and support up to one million tokens of context. Ultrafast generation — up to 4.5 times standard speed, per OpenAI — reaches $200-a-month Pro subscribers for the first time, starting with Sol. The launch lands in the same week OpenAI’s computer-use agents moved into hospital software — see our healthcare agents story — and days after Alibaba’s aggressive flagship pricing in the Qwen4-Max launch.

Why it matters

The frontier race is shifting from raw capability to capability per dollar. A flagship that scores higher while costing less per task — and a budget model that nearly matches it on real coding work — squeezes every rival pricing a premium model this month, Anthropic’s IPO-week lineup included. If Sol holds up on merge-ready code at one-fifth of Astra’s cost, expect it, not the flagship, to become the default engine inside a lot of agentic tools.

Source: OpenAI — GPT-6.2 Astra and Sol are now available (November 17, 2026).

Mohsen Daemi