Affiliate links on Android Authority may earn us a commission. Learn more.
OpenAI launches GPT-6.2 Astra and Sol, but not everyone is sold on the cheaper upgrade
Nov 18, 2026 — 4:11 AM ET

- OpenAI has launched GPT-6.2 Astra and Sol, promising improved performance and greater efficiency across professional and everyday tasks.
- Early testers reported problems with unproductive loops, premature task completion, and confusing responses despite impressive benchmark results.
- OpenAI believes some reviewers missed last-minute model improvements and acknowledges that there’s still room to improve the user experience.
Ever since OpenAI introduced GPT-6 Astra in September, much of its attention has been on delivering flagship-level intelligence at a fraction of the cost. We saw this with GPT-6.1 Sol and, more recently, GPT-6.1 Terra. Keeping costs down becomes particularly important as tools like ChatGPT Work take on entire projects rather than simply answering questions. While OpenAI’s latest models push that efficiency even further, some early testers have found that saving on tokens doesn’t necessarily save time.
OpenAI has introduced GPT-6.2 Astra and GPT-6.2 Sol, with the former becoming its most capable model for complex professional work, scientific research, and engineering. Beyond crunching numbers and writing code, Astra brings improvements that should be immediately visible in the things you create, from more thoughtfully designed websites and polished presentations to richer 3D designs.
Perhaps more impressive is that Astra achieves new benchmark highs across professional and scientific tasks at less than a third of its predecessor’s cost per task. Sol carries that emphasis on efficiency into everyday work, offering a more economical model for everything from writing to completing multi-step assignments.
But impressive benchmark scores don’t necessarily make for an impressive coworker. Among those given early access, a recurring frustration was the models’ tendency to get caught up in unproductive loops, abandon assignments prematurely, and produce summaries that were unnecessarily difficult to understand. In one particularly frustrating instance, days of attempted coding improvements reportedly left a project performing worse than before.
It wasn’t all bad, though. Early testers also praised Astra’s ability to identify problems other models missed, find unconventional solutions, and produce impressive 3D projects. The bigger issue appeared to be getting that intelligence to work consistently in their favor.
However, the models behind some of those reviews may not be quite the same ones you’re getting today. OpenAI’s Thibault Sottiaux explained on X that the company made substantial changes in the final three to four days before launch, which he believes some reviewers missed. Still, he acknowledged that improvements in creative taste and ease of use remain on the company’s agenda.
Both models are available in ChatGPT Work and Codex for Plus, Pro, Enterprise, and Edu subscribers, as well as through the API. OpenAI is also bringing Ultrafast access to $200 Pro subscribers for the first time, starting with Sol, while $500 Pro subscribers can use it with either model.


Conversation
Read what readers said or add your take.