Google is planning to release its next lightweight model, Gemini 3.2 Flash, at the I/O conference on May 20. The model's overall performance is roughly on par with GPT-5.5 but falls short of Anthropic's Mythos.
Abacus.AI CEO Bindu Reddy shared rumors that Gemini 3.2 Flash reaches 92% of GPT-5.5's performance on coding and reasoning tasks, yet costs just one-fifteenth to one-twentieth as much for inference, with most queries completing in under 200 milliseconds. She attributes this to Google's distillation and sparsification techniques, which essentially compress a frontier model down to Flash level without the typical performance cliff.
Gemini 3.2 Flash has been leaking for weeks. Earlier this month, traces appeared in an iOS app bundle and AI Studio metadata, followed by anonymous appearances on LM Arena. Early testers say the model shines at creative coding and even beats Gemini 3.1 Pro on some benchmarks.