GPT-6.2 Sol vs 6.1: Faster, Cheaper per Task, but Smarter?
Model
GPT-6.2 Sol vs GPT-6.1 Sol
What I did
I used both during my regular development work, paying attention to speed and output quality. This was hands-on use, not a timed A/B benchmark.
What happened
6.2 felt noticeably quicker, but I didn't see an equally obvious reasoning leap. Plotly independently measured 10.5s vs 17.6s median per question, with 158 vs 258 tool calls. The trade-off: 6.2 scored 40/43 while 6.1 scored 43/43. Same token pricing, but the full run cost 2.61.
Worth it?
Yes for everyday iteration speed. Not a guaranteed accuracy upgrade.