Google's Gemma 4 models are now available on Vercel AI Gateway. Two variants are offered: a 26B Mixture-of-Experts model (activating only 3.8B parameters during inference for lower latency) and a 31B Dense model (all parameters active, better for fine-tuning). Both support 256K context, 140+ languages, native vision, function-calling, agentic workflows, and structured JSON output. Access them via the AI SDK using model identifiers `google/gemma-4-26b-a4b-it` or `google/gemma-4-31b-it`, with no markup and no separate provider account required.
Sort: