Vercel AI Updates: August 6, 2026

1. Vercel Launches the New v0 API for Programmatic App Generation

Vercel. Vercel released a new v0 API that gives developers programmatic, headless access to v0’s app-building agent, letting them send prompts and receive generated applications with live previews. The API supports synchronous, asynchronous, and streaming responses, design system integration, and direct deployment to Vercel. It is now generally available. Source

2. Vercel AI Gateway Arrives on AWS Marketplace

Vercel. Vercel’s AI Gateway is now available on the AWS Marketplace, letting teams procure it through existing AWS accounts and consolidate inference spend onto their AWS bill. The gateway provides a single endpoint to hundreds of models with built-in reliability, cost controls, and governance, with no markup over direct provider pricing. Source

3. Vercel Adds OpenTelemetry Trace Export for AI Gateway via Drains

Vercel. Vercel now emits an OpenTelemetry trace for every AI Gateway request and can export it through Vercel Drains to any compatible endpoint. Traces capture model and provider routing, token usage, and performance metrics without prompt content, with sampling controls and pricing at $0.05 per 1,000 traces. The feature is available to Pro and Enterprise teams. Source

4. Meta’s Muse Spark 1.2 Coding Model Lands on AI Gateway

Vercel. Vercel added Meta’s Muse Spark 1.2 coding model to AI Gateway, an updated release positioned for stronger code generation and multi-round, iterative development. Developers can route to it through the gateway’s unified API alongside other supported models. Source

5. Vercel Sandbox Scales to 10,000 Concurrent Instances

Vercel. Vercel Sandbox now supports up to 10,000 concurrent sandboxes and 5,000 vCPUs per minute, up from a prior 150 vCPU baseline, with quotas that scale dynamically based on sustained usage. The increase targets teams running large fleets of ephemeral compute environments for agent and code-execution workloads. Source