“In testing by Artificial Analysis, the Gemini 3.7 Flash model ran at 340 tokens per second.”