Gemma 4, Google's latest open-weight multimodal model, is now running on Cerebras at ~2,300 tokens/sec, making it the first open-weight multimodal model to break the 2,000 tok/s barrier. The 31B dense model (30.7B-parameter) supports vision, reasoning, long context, chain-of-thought, function calling, and agentic capabilities, with a 131K context window and 140+ languages. According to Artificial Analysis, Gemma 4 represents a major jump over Gemma 3 26B and ranks among the strongest open-weight models in its Intelligence Index. In testing, a side-by-side document analysis of the 60-page DeepSeek-V4 technical report returned in 1.79 seconds on Cerebras, compared to almost 25 seconds on GPU providers, a 17× speedup. For image analysis, both inference providers looped over the exact same set of 80 images, batched the same way, with the same prompt, leaving speed as the only variable.