AI Gateway adds confidence-based decision fallbacks
AI Gateway now supports escalating requests to a fallback model when the primary model's confidence falls below a threshold you set. Configure confidence conditions using providerOptions.gateway.models to automatically route low-confidence responses to secondary models, with support for Choice, Score, and Boolean questions.
can now escalate a decision request to a fallback model when the primary model's answer trips a confidence condition you set.AI Gateway
Conditions can be combined, so an escalation can depend on more than one signal. The plain model names you already list in keep catching outright errors, so existing fallbacks are unaffected. Confidence conditions cover Choice and Score questions. Boolean questions escalate on a probability range instead.models
Set the fallback by adding one conditional object to . Requests without it keep their existing behavior.providerOptions.gateway.models
Leave out and the condition checks every question of the matching type, so escalates when any Choice or Score answer falls below 0.6.question{ "confidenceBelow": 0.6 }
Because a triggered fallback runs a second decision, it bills both stages.
Decision fallbacks are in beta. Learn more in the .decision fallbacks documentation
Source: original entry ↗
More from Vercel
Follow Vercel to get its new changes in your feed and email digest.
Gemini 4 Argon now available on AI Gateway
Google's Gemini 4 Argon model is now available through Vercel's AI Gateway with support for coding, agentic work, and complex reasoning tasks. The model features a 1M token context window, multimodal input capabilities (text, image, PDF, video), tool calling, and configurable thinking levels for enhanced performance.
MAI-Image-2.6 and MAI-Image-2.6-Flash now available on AI Gateway
Microsoft's MAI-Image-2.6 and MAI-Image-2.6-Flash models are now available on AI Gateway, enabling image generation and editing through a single API key without requiring separate provider accounts. Both models support photorealistic imagery, text rendering, targeted edits, and up to five reference images with output sizes up to 1536×1536.
Repository checkout and Git LFS downloads are now faster
Repository checkout and Git LFS downloads are now faster on Vercel, reducing build preparation time. Average checkout time improved by 20%, with LFS-enabled projects seeing a 54% improvement and saving approximately 13 seconds per deployment.