In short: Vercel AI Gateway is one shared entry point to many AI models. You send the request to it, and it passes the request on to the model of the right provider.
By Vercel's own description, it is a unified API for calling models that tracks usage and cost and lets you set retries and failover if a provider stumbles. You pick a model by name in the form provider/model. Change the name and you change the model, while the rest of your code stays the same. Vercel also lists Zero Data Retention support, budgets for API keys and routing rules.
The most important thing for your bill: according to Vercel, AI Gateway reflects the provider's pricing with no markup and charges no platform fee on inference. The same applies to Bring Your Own Key (BYOK), meaning when you use your own key with the provider. These are Vercel's words about its own product.
New: Qwen4-Max from Alibaba
On 15 November 2026, Vercel announced that Alibaba's Qwen4-Max is available in AI Gateway under the name alibaba/qwen4-max. According to Vercel, it is the first model in the Qwen4 series: a reasoning model with native vision and a context window of 1 million tokens. That includes Vercel's claims that it uses fewer tokens and writes more secure code than earlier Qwen Max models. The page shows no independent test. To try it, Vercel points to the command npx vercel ai-gateway setup and then picking the model in your agent.
Primary sources
Vercel changelog: Qwen4-Max now available on AI GatewayThe visual is generated code-art, no third-party images.