Vercel has made another move in the AI arms race. Its new AI Gateway is set to elevate its position beyond a frontend hosting platform.
The Gateway allows users to switch between around 100 AI models, removing the effort of managing API keys, rate limits, or provider accounts. As it’s built on top of Vercel’s AI SDK, it includes the native production features like structured outputs by default. Developers’ focus can rest on building applications, not underlying infrastructure details.
Since it supports load balancing and failover mechanisms, the Gateway promises high availability and performance even during provider downtimes or high-demand periods. It also includes the ability to switch between models based on performance and pricing. This means teams can optimize costs by selecting the most suitable models for their needs.
To begin using the AI Gateway, developers can install the necessary packages using the following command: pnpm install ai@alpha @vercel/ai-sdk-gateway
Vercel revealed it has additional plans for future features to improve the Gateway. These include load balancing and failover to automatically distribute requests across multiple providers, bring-your-own-key support, and flexible pay-as-you-go billing. Additionally, the company also plans to introduce unified logging, usage tracking, and observability, and an OpenAI-compatible API.
Read the full announcement on Vercel’s blog.
Need help optimizing your Vercel setup? Download our Ebook.
Subscribe to Frictionless, the weekly newsletter by our CEO, Chris Lojniewski, and stay on top of the news.
