Vercel

Vercel Daemons can now use HTTP liveness checks to catch a worker that's still running but unhealthy. Add a lightweight health endpoint and point livenessProbe at it. Vercel restarts the container after repeated failed checks.

Responses from 200 to 399 count as healthy; redirects aren't followed. The defaults are a check every 10 seconds, a 1-second timeout, and a restart after three consecutive failures.…

Vercel is moving Data Cache out of beta and into Runtime Cache. Next.js fetch caching and unstablecache switch over on every version without code changes, but previously unbilled reads and writes become billable after migration.

Rates start at $0.40 per million read units and $4.00 per million write units, varying by region. Each operation rounds up to an 8 KB unit. Hobby includes 1 million read units and 200,000 write units monthly across all projects.…

Vercel has added Geist Serif to its type family. It's a high-contrast display face with fine, saw-toothed serifs and sharp hairlines, intended for headlines, posters, and other large text.

The first release has Regular weight; the other weights are marked as coming soon. You can install it with npm i geist and import from geist/font/serif, or download OTF, WOFF2, and variable font files. It's licensed under the OFL. Google Fonts availability is still coming.

Original source

InclusionAI's Ling 3.1 Flash 1019 is available on Vercel AI Gateway, with free requests until November 3, 2026. Requests begin charging at the published rate on November 3.

The model has a 1M-token context window. Vercel describes it as built for coding, multi-step reasoning, and agents that use tools.

Use inclusionai/ling-3.1-flash-1019 as the model name. For Claude Code, Codex, Cursor, and other coding agents, install the latest Vercel CLI, run npx vercel ai-gateway setup, then select the…

Vercel is merging Data Cache into Runtime Cache. Next.js fetch caching and unstablecache carry over without code changes. Pro projects get separate storage allocations for each environment with 8x the previous space, and Enterprise projects get 10x.

The billing change matters: previously unbilled Data Cache reads and writes become charged usage after migration. Regional rates start at $0.40 per million read units and $4.00 per million write units.…

Vercel Daemons can now call private services in their deployment through service bindings.

Declare the target service and an environment variable on the daemon, then make HTTP requests to the URL it receives. Vercel handles internal routing, authentication, and TLS.

Bindings work one way: services can't bind to daemons. Each daemon instance keeps calling its original deployment's services after a promotion or rollback, until that instance stops.…

Vercel Firewall now lets you log or block AI crawlers separately from AI assistants on all plans.

Crawlers collect content for indexing and model training. Assistants fetch pages in response to user actions or search queries. You can block the former while allowing the latter to retrieve your content.

In your project's Security Firewall Rules, enable AI Bots under Bot Management and select its gear icon. Set each category's action, then review and publish the changes.…

Vercel Firewall now lets you log or block AI crawlers and AI assistants separately with its AI Bots managed ruleset. Crawlers are the bots that collect content for indexing and model training. Assistants are the ones that fetch a page in response to a user's action or search query.

That split means you can, for example, block AI crawlers while still letting assistants retrieve your content to answer people's questions. Vercel updates the rules as new bots are added.…

Vercel says its updated IP geolocation provides city-level data for nearly all requests and postal-code data for 90% of traffic. It's enabled for all projects with no code or configuration changes.

The existing geolocation headers and geolocation() helper in @vercel/functions return the same fields, with more frequent data updates.

Some visitors may now resolve to a different city or country, especially those using VPNs, hosting providers, or leased IP ranges.…

Mistral Large 4 on Vercel AI Gateway has updated weights and lower prices for input, cached input, and output tokens. Existing calls to mistral/mistral-large-4 use the updated version and pricing without a model ID change.

Mistral says the weights improve agentic coding, defensive cybersecurity, spreadsheet creation and editing, web research, and resistance to prompt injection, with greater token efficiency on coding tasks.…