Summary
In its September 4, 2026 weekly update, Vercel added per-user budget controls to its AI Gateway and expanded model availability, including Claude Fable 5.1, Gemini 3.8 Flash, and Qwen 3.8 Max, so teams can route across providers through one API key with spend tracking and request traces.
What changed
Vercel's AI Gateway gained per-user budget controls and added several models: Claude Fable 5.1, Qwen 3.8 Max 0902, Gemini 3.8 Flash, and Muse Spark 1.3, with temporary discounts on others such as MiniMax H3 and H3 Max at 50 percent off and GLM-5.3 via DigitalOcean. The gateway continues to offer one-key access, provider fallbacks, spend tracking, and request traces.
Why it matters
Per-user budgets turn the model gateway into a governance and cost-control surface, not just a routing convenience, which is key for enterprises exposing many models to many developers. Same-week addition of the latest frontier-class models shows gateways racing to be the neutral point where model choice and spend are managed.
Evidence excerpt
Claude Fable 5.1 released. Qwen 3.8 Max 0902 released. Gemini 3.8 Flash released. Per-user budget controls added.