Vercel Put Le Chonk in the Gateway the Same Day

Mistral launched Large 4 in the morning. By the changelog, Vercel already had it behind one key.

On 6 October 2026 AI Gateway added mistral/mistral-large-4. You do not need a Mistral account. Call it from the AI SDK, OpenAI-compatible Chat Completions, the Responses API, or the Anthropic Messages API. Context on the Gateway card: 524,288 tokens in, 262,144 out. Price: $0.68 / million input, $2.09 / million output, cached input $0.07. A 50% promotion was on the catalog row the same day. Promotions expire. The model ID will not.

Image: Vercel Changelog, Mistral Large 4 now available on AI Gateway, 6 October 2026.

The agent path is the point

npx vercel ai-gateway setup detects installed agents, mints a Gateway key, and writes config. In Claude Code you run /model and pick Large 4. fx signs in with Vercel or pastes a key, then /model mistral/mistral-large-4. That is how a lab preview becomes a Tuesday afternoon model switch instead of a procurement ticket.

Gateway is the right place to try a preview this large. You get routing, retries, and failover without baking Mistral into every service. You also get a second set of numbers: Vercel's context window is half of Mistral's million-token claim. If your eval assumes the Studio card, you will blow the Gateway ceiling. Pin the Gateway limits in the eval harness.

Use this week to run the same DeepSWE slice through Gateway that you already run through whoever you use for Opus. If le Chonk is cheaper per resolved issue, keep the route. If it is not, the one-key abstraction still paid for itself — you did not sign a second vendor contract to find out.

Sources

Vercel Changelog, Mistral Large 4 now available on AI Gateway, 6 October 2026. https://vercel.com/changelog/mistral-large-4-now-available-on-ai-gateway

Vercel AI Gateway, Mistral Large 4. https://vercel.com/ai-gateway/models/mistral-large-4