The Mini model scores 54.38% on SWE-bench Pro (a big jump from the last version), while Nano is a bit behind but remains fast.
Pricing is super wallet-friendly: Mini costs $0.75 per million input tokens ($4.50 per million output tokens), and Nano comes in even lower at $0.20 per million input tokens ($1.25 per million output tokens), way less than previous models.

Early testers are impressed: Aabhas Sharma from Hebbia says Mini beats competitors on tasks and citation recall without breaking the bank, even outperforming GPT-5.4 in some areas.
Abhisek Modi from Notion found it better than GPT-5.2 for formatting and tool use, hinting these models could be real game-changers if you want smart AI that won't drain your budget.