Vercel

Vercel announces general availability of Python support: CPython on Linux, automatic framework detection, Fluid compute, and idiomatic SDKs for storage, queues, durable workflows and AI models, with a shared frontend and backend deploy workflow.

Vercel's AI Gateway passes through upstream cuts on the GPT-5.6 line: Luna falls 80%, Terra 20%, across both short- and long-context tiers, while Sol holds price but its fast mode speeds up from 1.5x to 2.5x. No markup is added and model IDs stay the same.

Vercel's v0 can now install private npm packages using credentials stored as shared environment variables, letting teams pull internal design systems and component libraries directly into v0-generated projects without exposing secrets to the model.

Vercel's skills.sh registry reached one million published agent skills and roughly 280 million installs within seven months of Anthropic's Agent Skills launch, growing faster than GitHub, npm, or the App Store did to their first million.

Vercel removed the 100/500/1,000-store caps on Blob storage across Hobby, Pro, and Enterprise, and now bills store creation as an Advanced Operation ($5/million on Pro), so teams can split data per tenant, environment, or branch freely.

Vercel's dashboard now builds and deploys eve agents directly: write the instructions, pick any AI Gateway model, attach a Next.js chat or Slack surface, and connect tools through Linear, Notion or a custom MCP server. It scaffolds a private repo you can customize afterward.

AI SDK 7.0.150 adds experimental_streamSpeech, which streams audio from text deltas using speech models that implement the new streaming extension, with error handling, abort support and a mock model for tests.

A two-week public bounty program pays up to one million dollars for escaping a microVM agent sandbox, arguing that containment usually fails on an unaccounted network path rather than at the VM boundary.

Four Fish Audio speech models, three text-to-speech and one transcription, are now served on Vercel AI Gateway and free through September 18. Standard model names begin billing when the window closes, while a free-suffixed name stops serving instead.

Gemini 3.7 Flash is on Vercel's AI Gateway at half price until the end of 2026, with better issue resolution, less time stuck in failed agent loops, and code generated directly from design mocks.

Vercel published design.md, a single public file any coding agent can load to build pages that match its brand. It carries the brand, layout, and system guidance that previously only worked inside repositories where the product-design skill could read the design system directly.

Vercel's AI Gateway beta now escalates decision requests to a fallback model when the primary answer trips a confidence condition, such as a score below a threshold. Conditions can be combined, and triggered fallbacks bill both stages.

MiniMax M3 and M2.7 run free on AI Gateway through September 6 via GMI Cloud, either through the -free model IDs or by pinning gmicloud in provider order with fallback. The coding-agents setup command wires them into Claude Code, Codex, OpenCode, Cursor and others.

AI SDK 7.0.149 adds an interrupted finish reason for streams ending without a completion signal, enforces toolChoice 'none' before local execution, and processes only the first approval response per tool call.

Vercel Agent now runs inside Slack code channels, where a whole team can follow the work, direct it, and approve each change. It stays read only by default and records who requested and approved every action.

Vercel announces a temporary 90% price cut on DeepSeek V4 Flash when routed through Novita on its AI Gateway, available to Pro customers through August 11. Developers must set the model to deepseek/deepseek-v4-flash and pin Novita as the first provider in the gateway's order option to receive the discount; after the deadline the model remains available at standard rates.

Chat SDK's new Notion adapter lets a single agent answer comment threads on Notion pages alongside Slack, Discord and GitHub. Pages map to channels and comment threads to threads; mentions, edits, history and three attachments work, while buttons, modals and reactions do not.

The Vercel CLI now lets coding agents search, price-check, and initiate domain purchases using a bundled skill. Non-interactive buys return a machine-readable error so a human must confirm price, auto-renew, and contact details.

Grok Build joins Vercel's AI SDK harness layer through the same HarnessAgent interface as Claude Code, Codex, Deep Agents, OpenCode and Pi, so runtimes can be swapped without changing application code.

Vercel's v0 adds inline Marketplace integrations for Resend, Amazon OpenSearch, MongoDB Atlas, Algolia, and Clerk: prompting for a feature that needs a provider now renders a connect card, and v0 loads that provider's own agent skills.

OpenAI's GPT-Live 1, a full-duplex voice model that can listen and speak simultaneously, is now on Vercel AI Gateway with client delegation to hand off reasoning to a separate text model mid-session.

Vercel's September AI Gateway index: open-weight models hit 56% of token volume (up from 7% in December), prices fell for a third straight month, and GPT-6 Astra took twice Fable 5.1's spend share in its first two days while Opus 5 tripled its share.

AI Gateway adds asynchronous video generation: a webhook option, an SDK polling mode, and a startVideo/getVideoStatus pair that returns a job ID before rendering begins. All four paths cover text-, image- and reference-to-video; existing generateVideo calls behave as before.

Vercel AI Gateway now supports per-user spend limits, covering every API key and app token attributed to a person. Budgets stack with key, project and team limits, reset on a daily, weekly or monthly cadence, and alert by email at 50, 75 and 100 percent of allocation.