974 articles on Performance.
We migrated the Cloudflare Blog to EmDash to prove our stack at massive scale. Here is how we stress-tested performance, safely routed production traffic, and redesigned the frontend experience.
As AI transforms software economics, the standard revenue playbook is breaking down. Learn how leaders around the world are preparing for agent buyers, updating processes for faster pricing iteration, and building more flexible infrastructure.
Gisting compresses context into a set of learned tokens, preserving its quality while making the model faster and cheaper.
Learn how to implement high-performance personalization with Next.js and Vercel. Discover best practices, avoid common pitfalls, and deliver fast, scalable, and revenue-driving ecommerce experiences without sacrificing speed or user experience.
With AI, speed comes easy, but so does tech debt. That’s why the best teams don’t just ship something fast; they consider it carefully, build it efficiently, and make it stand out.
In July, we experienced eight incidents that resulted in degraded performance across GitHub services.
Agents can write code faster than teams can review, deploy, and maintain it. Today we’re introducing the Agent Development Lifecycle and the Cloudflare primitives that underpin it.
Next.js 16.3 applications on Vercel send 45% fewer prefetch requests, 17% fewer static assets, and have 2x faster path metadata serving.
Agents need more than just a container to scale. We're introducing @cloudflare/computer, an agent runtime that dynamically orchestrates between fast, efficient isolates and full Linux containers to give every agent a computer of its own.
Serving frontier models like Kimi and GLM means fighting for GPU memory. Here's how we quantize KV caches, compress model weights, and add integrity checks to serve them faster, cheaper, and safely.
How a branch-free loop and byte-space arithmetic let GitHub case-fold every byte of code search at >45 GiB/s on a single core.
This blog explores Data Lineage using SQL/PGQ in PostgreSQL 19, with real life examples that can enhance your learning.
Vercel introduces the inaugural cohort of Vercel Certified Solution Partners. These industry-leading teams share our commitment to creating a faster, more accessible, and innovative web for our customers.
Learn about how Vercel Functions help you run secure, highly available, and fast compute.
Subagents get justified by time saved and parallel execution, but that's not what matters most in long-running multi-agent work. Every token in the orchestrator's context is competing for its attention, and the real value of a subagent is what it keeps out of that context, not how fast it runs. I argue that subagents should be treated as a tool for protecting the orchestrator's working memory, off
We moved five high-traffic checkout extensions to remote-dom and Polaris web components, cutting bundle sizes drastically and making checkout faster.
LLMs generate code incredibly fast, but to ensure they generate exactly what is intended, they need clear boundaries. Abstractions and Domain-Specific Languages (DSLs) provide a strong harness that guides LLMs right from the start. The example of Tickloom - a domain model and DSL for illustrating distributed system behavior - shows how we can use an LLM as a partner to iteratively build a DSL and
The fast-growing customer experience platform explains how Figma MCP and Figma Make helped them scale a new design system and keep pace with customer requests.
With accurate first results and fast iterations, OpenAI's GPT-5.6 gives your builds a strong start in Figma Make.
Design leader Jen Dunnam offers her advice for this new age of AI—from building a strong team to how to think about speed.
In June, we experienced six incidents that resulted in degraded performance across GitHub services.
New Innovation Graph data shows global developer communities growing faster than ever, with collaboration reaching new highs across many economies.
We have been building websites inside boxes for years on WordPress. Let’s take a closer look at [Kirki](https://kirki.com/), the first freeform visual builder with an infinite canvas, and explore how it redefines the experience with cleaner performance, full design freedom, and zero plugin dependency.
Explore AI integration strategies with Vercel's AI SDK. Learn to choose providers, optimize performance, and future-proof your apps. Discover tools for seamless AI deployment and scalability.
Vercel AI Cloud combines unified model routing and failover, elastic cost-efficient compute that only bills for active CPU time, isolated execution for untrusted code, and workflow durability that survives restarts, deploys, and long pauses.
Vercel Ship 2025 added new building blocks for an AI era: Fast, flexible, and secure by default. Lower costs with Fluid's Active CPU pricing, Rolling Releases for safer deployments, invisible CAPTCHA with BotID. See these and more in our recap.
Fluid compute on Vercel combines serverless efficiency with server-like flexibility, reducing cold starts and cutting compute costs by up to 85%. Scale intelligently, minimize latency, and optimize performance with zero config.
Unpacking the core usage metrics of Edge Requests and Fast Data Transfer. Learn how Vercel handles network routing and data transfer—while giving you full control over performance and costs.
Cline scales its open source coding agent with Vercel AI Gateway, delivering global performance, transparent pricing, and enterprise reliability.
Learn about the differences between the image formats JPEG, PNG, WebP, and AVIF with regards to compression techniques, resolution capabilities, pixel density, and effects on user experience for optimal web performance.
The best v0 prompts include three things. What you're building, who uses it and when, and your design constraints. This guide walks through the framework with side-by-side tests showing faster generation times, less code, and better UX decisions.
WebStreams had too much overhead on the server. We built a faster implementation. See how we achieved 10-14x gains in Next.js rendering benchmarks.
See how we built the data-heavy Black Friday-Cyber Monday dashboard to be cost-efficient, fast, and accurate. Building a data-heavy, real-time dashboard with a good user experience comes with challenges. Let's walk through how we overcame them.
Fluid compute now uses Active CPU pricing. Only pay CPU rates when your function is actively computing. Building on existing Fluid gains, this brings additional savings of up to 90% for workloads like LLM calls, AI agents, or tasks with idle time.
Audio/voice support is live on AI Gateway, with realtime. Build realtime, low-latency voice agents with the AI SDK on Vercel AI Gateway, plus text to speech and speech to text behind one API key and your existing tooling.
This article shows that problems that too many tables in a single PostgreSQL database can cause and how to deal with these problems.
Teams can generate UI faster than ever, but they still have to guarantee that what they ship is usable, secure, and maintainable. Accessibility as an operational capability rather than a compliance checklist or end-of-project audit, and what that looks like in practice.
Vercel now runs any HTTP server straight from a Dockerfile. Bring a Rails, Django, Spring Boot, or Go app and Vercel builds, deploys, autoscales, and runs it on Fluid compute with Active CPU pricing.
Vercel Flags is a platform-native feature flag provider built into the Vercel developer platform, server-side by default with zero impact on page performance.
We're excited to share Rich Harris, the creator of Svelte, has joined Vercel to make the Web. Faster.
We rearchitected the layers panel with new computation and caching strategies, resulting in 30–50% faster interactions in some of the largest, most complex files.
In May, we experienced nine incidents that resulted in degraded performance across GitHub services.
Lead Product Designer at Dropbox Sara Clayton used Figma Weave to add dimension to an Ideal Customer Profile in under two minutes. Steal her template to make your users feel real.
In our post about Project Glasswing, we made the argument that the architecture around a vulnerability matters more than the speed of the patch. Here we walk through what that architecture looks like, the threats it defends against, and how we run it ourselves as Cloudflare's customer zero.
There are a bunch of JavaScript animation libraries out there, and you might have wondered whether there’s a performance cost compared to traditional CSS transitions and keyframe animations. In this blog post, we’ll compare the same animation across several different strategies and see the differences firsthand. There’s some interesting nuance here!
Cloudflare has integrated with Anthropic's Claude Managed Agents to provide a fast, isolated execution environment for autonomous code delivery. This means builders can scale agent workflows globally while strictly controlling access to private backends and easily customizing their agent’s tools and runtimes.
In April, we experienced 10 incidents that resulted in degraded performance across GitHub services.
How the GitHub Issues team used client-side caching, smart prefetching, and service workers to make navigation feel instant.
We investigated a bug where CUBIC's congestion window became pinned at its minimum floor, causing a performance to plummet. The fix involved correctly measuring idle periods to distinguish RTT wait times from actual application idleness.
Agents are changing your code faster than your team can follow. Now you can close that gap with new MCP skills, architecture layouts, and more in FigJam.
We fine-tuned Qwen3-32B into a tool-calling agent that generates Flow automations: 2.2x faster, 68% cheaper, with a weekly retraining flywheel.
We are launching Flagship, a native feature flag service built on Cloudflare’s global network to eliminate the latency of third-party providers. By using KV and Durable Objects, Flagship allows for sub-millisecond flag evaluation.
Running LLMs across Cloudflare’s network requires us to be smarter and more efficient about GPU memory bandwidth. That’s why we developed Unweight, a lossless inference-time compression system that achieves up to a 22% model footprint reduction, so that we can deliver faster and cheaper inference than ever before.
By migrating our request handling layer to a Rust-based architecture called FL2, Cloudflare has increased its performance lead to 60% of the world’s top networks. We use real-user measurements and TCP connection trimeans to ensure our data reflects the actual experience of people on the Internet.
We built a custom technology stack to run fast large language models on Cloudflare’s infrastructure. This post explores the engineering trade-offs and technical optimizations required to make high-performance AI inference accessible.
In March, we experienced four incidents that resulted in degraded performance across GitHub services.
We used DSPy to turn prompt engineering for our relevance judge into a measurable, automated optimization loop, improving task performance, cost, and how reliably it works in production.
Modern hardware is remarkably fast, but software often fails to leverage it. Mechanical sympathy - a concept borrowed from racing and popularized in software by Martin Thompson - is the practice of creating software that is sympathetic to its underlying hardware. This practice can be distilled into a set of everyday principles: Predictable memory access, awareness of cache lines, the single-writer
The path to better performance is often found in simplicity.
