AI BriefingVercel Blog news
The latest AI and developer news from Vercel Blog, with short summaries.
Visit website →Vercel Flags SDK cuts cold-start initialization time by 7x in version 1.5.3
The update eliminates network waits during initialization and stops usage reporting from extending function duration.
Vercel Firewall adds independent controls for AI crawlers and assistants
The new AI Bots managed ruleset is available on all Vercel plans and allows separate actions for indexing bots and user-facing assistants.
Vercel Plugin Launches in Google Antigravity Marketplace
The plugin provides agents with guidance on Vercel APIs and recommended patterns for products like Vercel Functions and Sandbox.
Vercel IP geolocation updates deliver city-level data for nearly all requests and postal codes for 90% of traffic
The update enables city-level resolution for nearly all requests and refreshes data more frequently to account for IP range reassignments.
Mistral Large 4 updated weights and lower pricing on AI Gateway
The update improves agentic coding, defensive cybersecurity, and token efficiency without requiring changes to the existing model ID.
Paseo Adds Vercel Sandbox Plugin for Cloud Agent Execution
The plugin allows users to start agents in the cloud from a laptop and control them via phone, saving files and history when the sandbox stops.
Vercel Introduces Virtual Models on AI Gateway for Centralized Model Management
The feature allows developers to update models and configurations via a single ID, supporting 125 new models added in the last 90 days.
Particle-1, a DeepSeek V4.1 Flash-based model, launches on AI Gateway
Post-trained on DeepSeek V4.1 Flash, the model targets software engineering tasks with shorter reasoning and fewer tokens.
Vercel Container Registry raises compressed image layer limit to 2 GB
The limit increases from 500 MB to accommodate large toolchains and dependency sets, while the total image size cap stays at 15 GB.
Vercel CLI 64.2.0 Allows Agents to Enable Web Analytics and Observability Plus
The update removes the need for an interactive terminal, allowing agents to run enablement commands from a linked project directory.
Vercel CI Beta runs only affected tasks, cutting CI time by 2-6x
The new system uses Turborepo to identify and run only the specific tasks affected by a code change, leveraging Fluid compute for faster feedback.
Box Mount enables agents in Vercel Sandbox to sync Box files via standard filesystem operations
The feature is in private preview, requiring users to request access from Box and unmount before stopping the sandbox to ensure data persistence.
T3 Code Adds Support for Open-Source Coding Agent Harness fx
The built-in provider supports Supervised and Auto modes using existing fx models and skills.
AI SDK fx adapter moves harness execution to host process
The rewrite uses libfx to eliminate sandbox bridges and port exposure while keeping model credentials on the host.
Vercel Plugin adds API guidance and specialist agents to OpenCode
The plugin requires OpenCode 2 and includes specialist agents for deployment, performance optimization, and AI application design.
Vercel Container Runtime Now Supports Plain Dockerfile Deployment
The beta runtime builds and routes traffic to images from root Dockerfiles or Containerfiles when no other framework is detected.
Vercel adds AI traffic insights to Observability
The feature reveals request counts by verified bots, most-crawled pages, and assistant retrievals, with CLI access requiring Observability Plus.
Vercel AI Gateway integrates 14 ElevenLabs audio models with 50% launch discount
The partnership covers six audio categories including speech-to-text, streaming TTS, realtime voice agents, music, sound effects, and multi-voice dialogue, accessible via a single API key.
Vercel Agent sessions cost 67% less due to rebuilt context management
The update improves prompt-cache reuse and incremental context handling, requiring no user action.
AI SDK HarnessAgent adds image input support for all harness adapters
The update allows passing images alongside text in the standard AI SDK message format, processed by underlying vision models.