ChainThink report, on May 20, Google will unveil its next-generation lightweight model, Gemini 3.2 Flash, at the I/O conference. According to Abacus.AI CEO Bindu Reddy, the model achieves 92% of GPT-5.5's performance in coding and reasoning tasks, with reasoning costs reduced to just one-fifteenth to one-twentieth of those for GPT-5.5, and most queries exhibit latency below 200 milliseconds. Overall performance is on par with GPT-5.5, though it clearly falls short of Anthropic's Mythos.

Bindu Reddy, CEO of Abacus.AI, further noted that Google’s distillation and sparsification techniques have compressed state-of-the-art models down to Flash-level efficiency without triggering the typical performance cliff.

Gemini 3.2 Flash had earlier shown signs of leakage, with traces appearing in iOS app build packages and AI Studio metadata in early May. It later surfaced anonymously on the LM Arena benchmark, where early testers reported standout performance in creative coding tasks, with some benchmark results surpassing those of Gemini 3.1 Pro.