What Vercel Shipped in Late October 2026.

Vercel’s late October releases cover AI Gateway models, Daemon controls, IP geolocation, Runtime Cache and a new typeface. The changes give developers more model options, private service connections, automatic health checks and larger cache allocations.

Mistral Large 4 Gets Updated Weights and Lower Pricing

AI Gateway now serves Mistral Large 4 with updated weights and lower input, cached input and output token pricing, according to the changelog entry. Existing users keep the mistral/mistral-large-4 model ID, which automatically points to the new version and pricing.

IP Geolocation Adds Higher Resolution

Vercel now provides city-level geolocation data for nearly all requests and postal-code data for 90% of traffic. The IP geolocation update is active for every project without code changes, using the existing headers and geolocation() helper in @vercel/functions.

Strix Connects to AI Gateway

Strix now supports AI Gateway, with usage and cost tracking across providers. The Strix integration entry shows how to configure a model through the STRIX_LLM environment variable using the vercel_ai_gateway/ prefix.

Vercel Daemons Gain Service Bindings

Vercel Daemons can now call private services in the same deployment through service bindings. As described in the service bindings entry, developers declare the target service and an environment variable, then use a standard HTTP request while Vercel handles routing, authentication and TLS. Bindings only work from daemons to services, not the other way round.

Vercel Daemons Add HTTP Liveness Checks

Developers can set a health endpoint for a Daemon using livenessProbe in vercel.json. Vercel periodically requests that endpoint and restarts the container if the check fails, according to the liveness checks entry.

Ling 3.1 Flash 1019 Joins AI Gateway

InclusionAI’s Ling 3.1 Flash 1019 is available through AI Gateway for free until 3 November 2026. The Ling release entry lists a 1M-token context window and support for coding, multi-step reasoning and tool-use agents through the model ID inclusionai/ling-3.1-flash-1019. Published rates apply after the free period.

Runtime Cache Integrates Data Cache

Data Cache is now part of Runtime Cache, letting applications reuse fetched or computed data across requests without a separate caching service. The Runtime Cache blog post says existing Next.js fetch caching and unstable_cache implementations need no code changes, while Pro and Enterprise projects receive larger storage allocations.

Data Cache Exits Beta

The migration changelog sets out the allocations and pricing behind the move to Runtime Cache. Pro projects receive eight times the previous Data Cache space per project and environment, while Enterprise projects receive ten times their previous storage. Pricing starts at $0.40 per million read units and $4.00 per million write units, with Hobby teams receiving 1 million read units and 200,000 write units free each month.

Geist Serif Joins the Geist Family

Geist Serif is a high-contrast display typeface designed for headlines, print and high-density screens. The Geist Serif announcement describes fine saw-toothed serifs and sharp hairlines, adding an editorial option alongside Geist Sans, Geist Mono and Geist Pixel.