Vercel's changelog announces that new Pro and Enterprise teams no longer receive a Standard build slot and will instead pay usage-based rates for Standard builds. On Pro, builds cost $0.014 per build minute and the monthly usage credit applies, while Enterprise rates follow each contract. The post is a billing notice with no technical implementation details.
Vercel engineering blog
Loading
Vercel Blob now supports time-to-live settings that automatically delete blobs a set number of days after upload, from 1 to 730 days. The post covers the ttlDays option across put, copy, multipart, and presigned uploads, the setTtl method for changing existing blobs, and the CLI flags. It also explains the billing model, where setting a TTL costs one Advanced Operation per started 60 days.
Microsoft's Decision-1.1 model is now available through Vercel AI Gateway and returns structured answers (boolean, choice, and score types) rather than free-form text, with calibrated probabilities that let applications set action thresholds. The post notes the model uses the same request format as Decision-1, shows an AI SDK example, and states that decision requests appear in logs, count toward budgets, and support fallbacks.
Vercel announces eve 1.0, an open-source framework for production AI agents, and details the engineering behind it. Median hosted turn latency fell from 3.04s to 414ms through a single-orchestrator workflow design and WebSocket event writes, and estimated workflow cost for a 50-turn session dropped about 76% by storing only changed conversation data. The post also covers OpenTelemetry-based tracing, multi-tenant agent configuration, and dynamic tool loading.
Microsoft's MAI-Image-2.7 and MAI-Image-2.7-Flash are now in public preview on Vercel AI Gateway, with maximum output raised to 2048 x 2048 pixels and transparent background support. The post shows how to generate and edit images through the AI SDK and AI CLI using a single gateway API key, including the providerOptions.azure.background setting and up to five reference images.
Vercel added an Ignored Build Step preset that skips preview builds for branches without an open pull request, while production and pull-request branches still build. On GitHub and GitLab projects, skipped previews are canceled before consuming Build CPU Minutes, and the preset fills in the ignore command automatically.
Vercel's AI Gateway now supports OpenAI's GPT-6.2 Astra and Sol models, accessible through the model IDs openai/gpt-6.2-astra and openai/gpt-6.2-sol. The announcement covers opting into the Ultrafast service tier via providerOptions.openai.serviceTier, using the models in coding agents through the Vercel CLI setup command, and a 60% lower Ultrafast output-token price compared with GPT-6 Astra.
The post argues that software used by AI agents needs evals, meaning repeated, graded runs rather than single tests, and lays out an agentic testing pyramid. It walks through building an end-to-end eval with Harbor on Vercel Sandbox and AI Gateway, including choosing between pass@k and pass^k, designing tasks and verifiers, and analyzing traces to find failure causes.
Vercel's Spend Management now supports budgets for individual projects and products, alongside team budgets. Project budgets can optionally pause a project's production deployments at 100% without affecting other projects, and alerts fire at 50%, 75%, and 100%. The post also covers CLI commands for listing and setting budgets, and notes that only team and project budgets support automatic pausing.
Vercel's AI Gateway now offers ByteDance's Seedance 2.5 Flash, a faster and lower-cost video generation model that produces synchronized audio. The post shows how to call the model through the AI SDK with text prompts and with image and video reference inputs, which are numbered in the prompt, and notes supported clip lengths, resolutions, and aspect ratios.
Vercel's AI Gateway adds Private Inference, letting enterprises connect fine-tuned, self-hosted, or prerelease model endpoints alongside catalog models under one API key, SDK, and logging pipeline. The post covers routing by model ID, team-scoped access with encrypted credentials and optional Zero Data Retention, fallback to catalog models, and a CLI and code example for registering a private provider. The announcement is mostly a product feature rundown with some concrete configuration detail,...
Tencent's Hy4 Flash is now available on Vercel AI Gateway with a 50% introductory discount through November 29, 2026. The model keeps the Hy4 feature set, including preserved reasoning, tool calling, structured outputs, prompt caching, image input, and a 1M-token context window, and reasoning can be disabled for high-volume workloads. The post includes a usage snippet using the model name tencent/hy4-flash.
Vercel's Instant deployments now accept static sites that include CSS, images, fonts, and other assets, not just HTML and Markdown. Eligible deployments skip the build step and go live in about a second via the CLI or REST API, subject to limits of 100 files and 25 MB total. Deployments with JavaScript, framework presets, custom build settings, or Git triggers still build as before.
Vercel released version 1.0 of its Python SDK for Vercel Sandbox, stabilizing the public API for running commands and managing files in isolated Linux microVMs. The release adds context managers for automatic cleanup, AnyIO-based async support for asyncio and Trio, and a sync module, and it switches command input and output to bytes, a breaking change for 0.x users.
Vercel's changelog announces that Alibaba's Qwen4-Max model is available through AI Gateway using the model name alibaba/qwen4-max. The post covers the model's reasoning, vision, and 1M-token context capabilities, plus a basic AI SDK usage snippet. Most of the content is product availability and vendor-claimed capabilities rather than transferable engineering detail.
Vercel announces that Parallel Decisions Fast is available through its AI Gateway at no cost until November 23, 2026. The model searches the live web to answer boolean, choice, and score questions with probabilities and cited sources, and it can be called through the AI SDK or an OpenAI-compatible API using a single gateway key.
Vercel's Token Pass add-on now includes $150 in monthly credit, up from $80, at the same $30 per team per month. The credit covers featured open-weight models served through AI Gateway and resets each billing period without rollover. The post also describes how requests fall back to the team balance or stop when the credit runs out.
Workflow SDK 5 introduces end-to-end AES-256-GCM payload encryption with per-run keys, regional run placement, and a rewritten engine that reduces queue hops and replay overhead. The post includes benchmark results and methodology comparing v4.8.12 with v5.2.0, along with WebSocket streaming and payload compression details.
Vercel Security Review now scans code changes for vulnerabilities in GitHub repositories that are not linked to a Vercel project. By default it covers all unlinked repositories the Vercel GitHub App can access, and teams can limit coverage to selected accounts or repositories in Agent settings. Each push to a covered repository triggers a review.
Vercel has raised the build cache limit on all build machines to 5 GB, up from 1.5 GB on Basic and Standard, 3 GB on Enhanced, and 4.75 GB on Turbo. Projects can now keep more installed dependencies and framework files between deployments, which can shorten install and build times. The change applies automatically at no extra cost and requires no configuration.