Gemma 4 on AI Gateway
Google's Gemma 4 is now live on Vercel's AI Gateway—two open models (26B MoE, 31B Dense) with 256K context, vision, and function-calling built in.
Why it matters
Vercel democratizes access to Google's latest open LLMs through unified API infrastructure, lowering barriers for developers to integrate production-grade models without managing multiple provider integrations.
The key facts
7 to knowGemma 4 26B (MoE) activates only 3.8B parameters during inference for lower latency
Gemma 4 31B (Dense) with all parameters active for higher output quality
256K context window support across both models
140+ language support with native vision capabilities
Function-calling and agentic workflow support
Available via Vercel AI Gateway with unified API, cost tracking, failover, and intelligent provider routing
Models built on Gemini 3 architecture
Go to the source
Vercel Blogvercel.com
Publisher excerpt: Gemma 4 26B (MoE) and 31B (Dense) from Google are now available on .Vercel AI Gateway Built on the same architecture as Gemini 3, both open models support function-calling, agentic workflows, structured JSON output, and system instructions. Both support up to 256K context, 140+ languages, and…