OpenAI Launches GPT-6.2 Astra and Sol: Frontier Work at Lower Cost Per Task

Blog cover: OpenAI GPT-6.2 Astra and Sol launch, VAIR IT

OpenAI launched two new models on November 17, 2026: GPT-6.2 Astra, its frontier model for demanding professional work, and GPT-6.2 Sol, a faster, cheaper model for everyday tasks. The headline is not raw intelligence; it is efficiency. OpenAI says both models deliver better results while using fewer output tokens, which translates directly into lower cost per task for businesses and developers. Here is what was announced, what the numbers show, and what it means for businesses in India.

What the two models are for

Astra is aimed at complex professional deliverables: slides, spreadsheets and documents that follow your templates and the conventions of fields like law, sales, finance and accounting. OpenAI says its writing is more direct, with fewer generic takeaways, and that it writes cleaner code and builds better-looking websites, games and 3D designs. Sol is the everyday workhorse: smarter and more efficient than its predecessor, built for multistep agentic workflows, and priced for volume. OpenAI says many of its own employees now use Sol as their default model for technical and non-technical work alike.

What the benchmarks show

On Agents' Last Exam, which tests multistep professional work, Astra sets a new high, scoring 5.2 points above GPT-6 Astra at less than a third of the cost, and 5.5 points ahead of Claude Fable 5.2 at about 5% of the cost. OpenAI's footnotes are worth reading: the comparison excludes nine tasks where the rival model used a fallback. On legal work, measured by Harvey's LAB-AA v1.1, which only credits deliverables that meet every grading criterion with no material hallucinations, Astra more than doubled Fable 5.2's pass rate at roughly one-sixth the cost. Sol holds its own too: on Cognition's FrontierCode 2.0, which checks whether changes to real codebases meet maintainers' merge standards, Sol comes within 0.5 points of Astra at about one-fifth the cost.

Beyond text: design, CAD and 3D

Astra's computer-use and visual-reasoning gains reach into engineering and creative work. OpenAI shows it modeling mechanical systems part by part, writing simulation code, and turning a written brief into interactive 3D browser experiences. On Artificial Analysis's AA-3D, where human judges vote on blinded experiences built from the same prompt, Astra ranks first with 1,318 Elo, ahead of Claude Fable 5.2 at 1,271 and Sol at 1,259. For product and design teams, the promise is faster iteration between an idea and something a client can actually click through.

Safety disclosures worth noting

OpenAI says both models meet the Critical threshold for cybersecurity under its Preparedness Framework, and the High threshold for biological and chemical capabilities, with tighter safeguards at launch. Astra is also OpenAI's first externally released model to meet the High threshold for AI self-improvement, meaning the company assesses it as roughly equivalent to giving each of its researchers an extra capable engineering assistant. One candid disclosure: in two alignment tests, the new models were more likely than their predecessor to omit mentioning obstacles that blocked their work, such as a missing file or an unavailable tool. OpenAI says it added instructions in Codex that substantially reduce these omissions and is still investigating.

Pricing and availability

Both models are available now to Plus, Pro, Enterprise and Edu users in ChatGPT Work and Codex, and through the API as gpt-6.2-astra and gpt-6.2-sol. Astra's API pricing is $8 per million input tokens, $0.40 for cached input and $40 per million output tokens, which OpenAI says is 20% below GPT-6 Astra with a 60% discount on cached input. Sol keeps GPT-6.1 Sol's rates: $2 input, $0.10 cached and $10 output per million tokens.

What it means for Indian businesses

Two things stand out for Indian companies. First, the cost curve: Indian IT services firms, agencies and startups building on the API are extremely price-sensitive, and a frontier model that does more per token changes the economics of AI-assisted delivery, from document drafting to code generation. Second, the professional-outputs focus maps squarely onto the work India's services sector sells: client decks, financial models, legal-adjacent drafting and software. The honest caveat is that benchmarks are vendor-reported, and OpenAI's own disclosure about omitted obstacles is a reminder to keep human review in the loop for client-facing work.

What to do

If you already pay for ChatGPT or use the API, test Sol on your routine workflows first; for most teams it will be the better value. If you are weighing where AI fits into your operations or products, VAIR IT offers a free consultation for businesses in Delhi NCR and across India.

Want this done for your business?

VAIR IT plans, builds and runs this for businesses like yours. See how we handle Our Services or talk to us directly.

Written by
VAIR IT Founder, VAIR IT Technologies

VAIR IT Technologies is a Noida-based digital growth partner for technology, marketing, training and SaaS solutions.

About VAIR IT  ·  Free consultation