Vercel AI Updates: June 25, 2026

1. GLM 5.2 Fast Arrives on Vercel AI Gateway via Wafer

Vercel. GLM 5.2 Fast is now available through Vercel’s AI Gateway routed over Wafer infrastructure, which Vercel says delivers twice the throughput compared to other serverless providers serving the same model. Performance benchmarks show 170+ tokens per second in small-context scenarios and 200+ in large-context cases, making it one of the faster paths to the GLM model family. Developers can access it in the AI SDK using the model identifier zai/glm-5.2-fast. Source

2. Vercel Now Deploys Node.js Servers with Zero Configuration

Vercel. Vercel’s platform now automatically detects a server.ts file and deploys it as a Node.js application without any additional configuration, supporting frameworks like Express and Koa alongside existing first-party frameworks. The deployment runs on Vercel’s Fluid compute infrastructure, removing the manual setup previously needed to host plain Node.js backends. The change closes a gap for teams that want to deploy lightweight API servers or backend services on Vercel without adopting a full framework. Source

3. Vercel OIDC Tokens Now Support Custom Audience Claims

Vercel. Vercel’s OIDC issuer now accepts custom audience claims, allowing deployments to request provider-specific tokens scoped to a particular service. The feature prevents token replay attacks across different services by ensuring exchanged tokens carry the correct audience while preserving the original claims and including an auditable delegation chain. Teams using OIDC for cloud provider authentication can now tighten their token scope without changing their existing identity setup. Source