Grok 4.5 from xAI is now accessible via Vercel AI Gateway, callable with model ID xai/grok-4.5 in the AI SDK. The model accepts text and image inputs and targets coding, STEM, and knowledge work.
Three reasoning levels, low, medium, and high, are available via the reasoning parameter, with high set as default. That tunability matters: you control the latency-versus-depth tradeoff directly, which is relevant for production workloads where inference cost compounds fast.
AI Gateway routes the call through a unified API with built-in failover, routing rules, per-key budgets, Zero Data Retention support, and custom reporting. No platform markup on inference, including BYOK requests. The full changelog details how the gateway handles uptime above provider-level guarantees, which is the part worth reading.
[READ ORIGINAL →]