November 18, 2026
OpenAI's GPT-6.2 Astra and Sol Are Here — And the Price Drop Changes Everything for Small Businesses
What if the most capable AI model available to your business just got 20% cheaper overnight? That is exactly what happened when OpenAI quietly dropped two new models, GPT-6.2 Astra and GPT-6.2 Sol…
OpenAI's GPT-6.2 Astra and Sol Are Here — And the Price Drop Changes Everything for Small Businesses
What if the most capable AI model available to your business just got 20% cheaper overnight? That is exactly what happened when OpenAI quietly dropped two new models, GPT-6.2 Astra and GPT-6.2 Sol, bringing frontier-level AI performance to a price point that makes serious business use not just possible but practical.
OpenAI released GPT-6.2 Astra and GPT-6.2 Sol on November 17, 2026, with both models available to Plus, Pro, Enterprise, and Edu users through ChatGPT Work and Codex. Astra is built for demanding professional work in fields like science, law, finance, and accounting, while Sol is designed for fast, versatile everyday tasks. API pricing for Astra dropped 20% from its predecessor, landing at $8 per million input tokens and $40 per million output tokens. Sol holds its predecessor's pricing at $2 input and $10 output per million tokens. For developers and businesses running AI-powered workflows, those numbers matter enormously at scale.
The benchmark story is striking. In the independent Artificial Analysis Intelligence Index, which rolls up nine evaluations across knowledge work, coding, science, and general knowledge, GPT-6.2 Astra ranks second globally with 62 points, sitting just behind Anthropic's Claude Fable 5.2 at 65. On the professional workflow benchmark Agents' Last Exam, Astra scores 5.5 points above Claude Fable 5.2 at just 5% of the cost per task. On the legal benchmark Harvey LAB-AA, which tests contract clause analysis and similar professional work, Astra more than doubles Claude Fable 5.2's pass rate at roughly one-sixth of the cost. Sol, the everyday model, nearly matches the original GPT-6 Astra on the FrontierCode 2.0 coding benchmark at only about 15% of that model's cost per task. According to Artificial Analysis, OpenAI dominates the Pareto frontier of cost versus performance, meaning no competing model is both cheaper and more capable across this range.
For small and mid-size business owners, the most immediately relevant development is not which model wins a benchmark. It is what this cost compression means for how you deploy AI in your day-to-day operations. When a model capable of drafting polished presentations, analyzing contracts, running agentic coding tasks, and handling complex financial research drops 20% in price while improving in performance, the return on investment on AI-powered workflows improves significantly. Teams that were previously priced out of using the most capable models for client-facing work, internal reports, or automated research pipelines now have genuine access to near-frontier capability at a fraction of yesterday's cost.
Sol specifically opens a door that matters for businesses without large technical teams. At $2 per million input tokens, Sol is the cheapest model in the top 10 of the Vals Index at $3.18 per task, according to Vals AI. It trails Astra by just 0.5 points on the FrontierCode 2.0 coding benchmark. For a small business owner who needs fast, competent output on a wide variety of tasks, from drafting emails to summarizing market research to building light automation, Sol delivers close to the best performance available at a price that fits a real-world budget. This is not a theoretical capability. It is a practical, accessible tool.
There is one important limitation to flag honestly. Both Astra and Sol showed notable weaknesses in legal research and general legal benchmarks in independent testing by Vals AI. On the Legal Research benchmark, Astra ranked 30th and Sol 52nd. OpenAI does offer a specialized product called Astra for Law that performs significantly better in that category, but the base models are not yet your best option for core legal research workflows. Additionally, Epoch AI's real-world testing found that Astra sometimes played it too safe on research tasks, skipping key estimates or missing the central question on complex projects. That means these models are powerful assistants, not autonomous decision-makers. The people on your team still need to review and direct the work.
This week, run a focused test with GPT-6.2 Sol on three recurring tasks your team handles manually right now: drafting a client-facing document, summarizing a report or research brief, and generating a first draft of a proposal or marketing asset. Use the actual ChatGPT Work interface or the API if your team has developer resources. Time both the AI-assisted version and the manual version, and compare quality. The 20% price drop on Astra and the accessible pricing on Sol make this week the right moment to find out exactly where these models save your team real hours, not hypothetically, but on actual work you are already doing.
The AI tools your competitors will be using in six months are available right now, at lower prices than they were last week. The gap between businesses using frontier AI and those still experimenting with it is getting narrower, but only for the ones who start testing today.
Originally inspired by: OpenAI launches GPT-6.2 Astra and Sol with lower API prices and some benchmark gains (https://the-decoder.com/openai-launches-gpt-6-2-astra-and-sol-with-lower-api-prices-and-some-benchmark-gains/)
Your turn
What is your traffic actually doing?
Send us your details and we will come back with a short, specific read on what your traffic, your pages and your pipeline are doing today — and the first three things we would change. A real person reads every submission, and you get the read whether or not we ever work together.
