Gemma 4 on AI Gateway
By Steven Van ·
Two open Gemma 4 models, a 26B mixture-of-experts and a 31B dense version, join Vercel AI Gateway with 256K context and native vision.
Vercel's AI Gateway now offers two open Gemma 4 models from Google: a 26B mixture-of-experts model and a 31B dense model. Both are built on the same architecture as Gemini 3 and support function-calling, agentic workflows, structured JSON output, and system instructions, with up to 256K context, 140+ languages, and native vision.
- 26B (MoE) activates only 3.8B of its 26B total parameters during inference, aimed at lower latency and faster tokens-per-second.
- 31B (Dense) keeps all parameters active during inference, targeting higher output quality and better suited as a foundation for fine-tuning.
To use them, set the model to google/gemma-4-31b-it or google/gemma-4-26b-a4b-it in the AI SDK. See the announcement for more.
Vercel
The platform for frontend developers — deploy, preview, and scale web apps and AI agents with zero config.
View Vercel →