HeadsUpAI

Vercel brings Google Gemma 4 to AI Gateway for high-performance agentic workflows

· Updated

Vercel integrated Google's new Gemma 4 family into its AI Gateway, providing immediate access to the 31B and 26B models. These models support a 256K context window and native multimodality. The 26B variant uses a Mixture of Experts architecture that activates only 3.8B parameters per token to prioritize speed.

This release bridges the gap between open-weight flexibility and the reliable tool-use required for autonomous agents. By supporting native function calling and structured JSON output, Gemma 4 enables complex reasoning tasks that previously required closed models. It offers digital sovereignty under an Apache 2.0 license without sacrificing intelligence-per-parameter.

You can call these models using the Vercel AI SDK with the google/gemma-4-31b-it or google/gemma-4-26b-a4b-it identifiers. The AI Gateway handles the infrastructure layer, including usage tracking, automatic retries, and failover logic. This setup is ideal for building local-first assistants or global applications that process large repositories.

Share this update