OpenAI launches GPT-6.2 Astra and Sol, cuts Astra API price 20 percent to $8 per million input tokens
Third-party tests show gains in science and cost efficiency, but flat coding scores, weak legal results and high hallucination rates on hard factual questions.
What happened
OpenAI released two new AI models, GPT-6.2 Astra for demanding professional work and GPT-6.2 Sol for everyday use. OpenAI says both beat their predecessors on benchmarks, which are standard tests for AI, and cost less to run. They are available in ChatGPT and through the API, the tool developers use to build apps.
Why it matters to you
Cheaper models make it less costly for companies to use AI for tasks like slides, research and code. Independent testers also found real weak spots, so results vary by task.
The takeaway
The new models are cheaper and strong in science, but not better at everything, so check them on your own work.
What it means for you
Coding scores are flat per Artificial Analysis, and Sol dropped three points on the Coding Agent Index, so test both models on your own codebase before switching.
Sol costs $3.18 per task against Astra's $10.68 in Vals AI testing, so Sol may be the cheaper default for routine work, though legal tasks ranked poorly.
Artificial Analysis says OpenAI models dominate the cost versus performance frontier, but Anthropic's Claude Fable 5.2 still leads its Intelligence Index.