Gemini 3 Flash scored 33.7% without tool use on Humanity's Last Exam.
Google makes Gemini 3 Flash the default model in the Gemini app and in Google Search's AI mode.
Google releases Gemini 3 Flash, a fast and lower-cost model derived from Gemini 3.
Day-0 inference support is provided in SGLang, the model is available on OpenRouter, and community H200 benchmarks report ~50K tokens/sec prefill and 5K–15K tokens/sec decode depending on config.
Xiaomi and community commentary claim MiMo-V2-Flash performs strongly on reasoning, coding, and agentic benchmarks and can compete with K2/DSV3.2 while using fewer parameters.