Image CDN in 2026: cdnjs vs jsDelivr vs Cloudflare Images

A single unoptimized hero image can add 400+ ms to LCP on a median mobile connection. Multiply that across a catalog of thousands of product shots or editorial photos, and your image CDN choice stops being a convenience and starts being an architectural decision with direct revenue impact. Yet most comparison articles still lump cdnjs, jsDelivr, and Cloudflare Images into the same bucket as if they solve the same problem. They do not. cdnjs and jsDelivr are static asset CDNs that happen to serve image files; Cloudflare Images is an image-specific processing and delivery pipeline. This article gives you a 2026-current technical comparison, a workload-profile decision matrix you will not find elsewhere, and concrete thresholds for when you outgrow free tiers entirely.

What Changed for Image CDN Architectures in 2026

Three shifts matter for anyone re-evaluating their image delivery CDN this year.

First, jsDelivr completed its migration to its own Rust-based edge worker layer in late 2025, replacing the previous multi-CDN failover between Cloudflare and Fastly with a unified routing fabric. As of Q1 2026, jsDelivr reports a global P50 TTFB of roughly 18 ms for cached assets, a measurable improvement over the 25-30 ms range it posted through most of 2024.

Second, Cloudflare Images expanded its free-tier variant count from 20 to 100 named variants per account in January 2026, and added AVIF auto-negotiation alongside WebP. The paid tier remains at $5 per 100,000 original images stored and $1 per 100,000 unique transformations delivered.

Third, cdnjs continues to operate as a Cloudflare-sponsored open-source project indexing libraries from npm and GitHub. Its catalog as of May 2026 lists roughly 5,300 libraries. It remains strictly a library CDN; there is no image transformation, no origin-pull, no custom asset hosting.

cdnjs vs jsDelivr vs Cloudflare Images: 2026 Technical Comparison

The key takeaway: cdnjs and jsDelivr are static asset CDNs. They serve files you or a package maintainer have already optimized. Cloudflare Images is an image delivery CDN with a processing pipeline. Comparing them head-to-head without acknowledging this architectural difference leads to the wrong choice.

Is jsDelivr Good for Serving Images from GitHub?

Yes, with caveats. jsDelivr can serve any file from a public GitHub repository or npm package via its CDN. Engineers commonly use this to host sprites, icons, and static marketing images for open-source docs sites. The URL pattern is straightforward and supports version pinning via tags or commit SHAs.

The limitation is that you get no transformation. Every variant you need (2x retina, mobile crop, WebP conversion) must be pre-generated and committed to the repo. For a docs site with 50 images, this is fine. For an e-commerce catalog with 10,000 SKUs at 4 variants each, you are now maintaining 40,000 pre-rendered files in Git. That workflow breaks down fast.

jsDelivr also enforces a fair-use bandwidth policy. There is no published hard cap, but sustained delivery above roughly 100 TB/month on a single package has historically triggered rate limiting or manual review. If your image traffic is predictable and moderate, jsDelivr is a legitimate free image CDN for GitHub-hosted assets. If it is not, you need a purpose-built system.

Can Cloudflare Images Resize and Optimize Images Automatically?

Yes, and this is its primary differentiator. You define named variants (e.g., "thumbnail", "hero", "og-card") with specific width, height, fit, and quality parameters. When a request hits a variant URL, Cloudflare generates the derivative on the first request, caches it at the edge, and serves subsequent requests from cache. As of 2026, AVIF is auto-negotiated based on Accept headers alongside WebP, so you do not need separate variant definitions per format.

The 100-variant limit on the free tier (up from 20 in 2025) covers most use cases. The real constraint is the 5,000 stored originals limit. Beyond that, pricing is $5 per 100,000 images stored and $1 per 100,000 transformations delivered. For a mid-size e-commerce site with 50,000 products, that works out to roughly $2.50/month in storage plus transformation costs that depend on cache hit ratio.

Workload-Profile Decision Matrix

This matrix maps workload characteristics to the image CDN option that fits best as of 2026. It is designed for architects evaluating tradeoffs rather than feature lists.

The pattern is clear: free tiers work for the top two rows. The bottom three rows need either Cloudflare Images on a paid plan or a dedicated image delivery CDN with volume pricing and real cache control.

When You Outgrow Free Image CDN Tiers

The breakpoint usually arrives between 5 TB and 25 TB of monthly image bandwidth. At that volume, jsDelivr's fair-use policy becomes a risk, and Cloudflare Images' per-transformation pricing starts to add up for high-variant catalogs. You also start needing granular cache purge, origin shield configuration, real-time analytics on cache hit ratios by region, and the ability to tune TTLs per path prefix rather than per account.

For teams that hit this threshold, BlazingCDN is worth evaluating. It delivers the stability and fault tolerance you would expect from CloudFront while pricing at a fraction of what AWS charges at comparable volume: starting at $4 per TB at the 25 TB tier and scaling down to $2 per TB at 2 PB. Flexible configuration and 100% uptime SLA make it a practical fit for image-heavy workloads where you need to scale delivery fast without renegotiating contracts. Sony is among its enterprise clients.

Operational Best Practices for Image CDN Delivery in 2026

• Pre-generate critical variants at build time. Even with on-the-fly resize, cold-start latency on the first request to a new variant can spike TTFB by 200-500 ms. Warm your cache for above-the-fold images during deployment.
• Instrument cache hit ratio per edge region. A global average of 95% hides the fact that your Singapore edge might be at 60% because of low traffic volume and short TTLs. Monitor per-region.
• Pin versions for library assets on cdnjs and jsDelivr. Unpinned "latest" URLs cause cache misses on every release and create unpredictable behavior in production.
• Set immutable cache headers for versioned image paths. If your image URLs include a content hash or version string, cache-control: public, max-age=31536000, immutable eliminates conditional requests entirely.
• Measure LCP attribution, not just CDN TTFB. Your CDN might report 15 ms TTFB while your real-user LCP sits at 2.8 seconds because the image is render-blocked or lazy-loaded above the fold.

FAQ

Is cdnjs a true image CDN?

No. cdnjs is a library CDN that indexes open-source JavaScript and CSS packages. It can serve image files that are part of those packages (icon sets, for example), but it does not support custom image uploads, on-the-fly transformations, or format negotiation. If your primary need is image delivery, cdnjs is the wrong tool.

Can I use jsDelivr as a free image CDN for a production e-commerce site?

It is technically possible but inadvisable at scale. jsDelivr's fair-use policy is not designed for high-bandwidth commercial image delivery. You also lose format auto-negotiation, resize, and crop capabilities. For anything beyond a small catalog, a dedicated image optimization CDN or Cloudflare Images is a better fit.

How does Cloudflare Images pricing compare to running my own image proxy?

At low volume (under 100,000 transformations per month), Cloudflare Images is almost certainly cheaper than self-hosting an imgproxy or Thumbor instance plus the compute and bandwidth to run it. At high volume (millions of transformations), the per-delivery cost adds up, and a self-managed proxy behind a CDN with volume-based pricing can be more economical. The crossover point depends on your cache hit ratio and variant count.

What is the best free image CDN for a static site with under 1,000 images?

jsDelivr serving pre-optimized images from a GitHub repo is the simplest zero-cost option. Generate your variants at build time using a tool like sharp, commit them, and reference the jsDelivr CDN URLs. For sites already on Cloudflare, the free Images tier (5,000 originals, 100 variants) handles this use case with the added benefit of automatic AVIF/WebP negotiation.

Does switching image CDN providers affect my Core Web Vitals?

It can, significantly. LCP is directly influenced by image delivery latency. Switching from a CDN with poor regional coverage to one with a denser edge footprint can reduce LCP by hundreds of milliseconds. Conversely, a migration that breaks cache warming or changes image dimensions without updating width/height attributes can cause CLS regressions. Measure before and after using CrUX data, not lab tests alone.

Your Move This Week

Pull your real-user LCP data from CrUX for the last 28 days. Filter to your highest-traffic page templates. For every template where LCP is an image element, trace the request waterfall: DNS, connection, TTFB, download. Identify whether your bottleneck is edge latency, origin response time, or image weight. That data will tell you whether you need a faster CDN, an image processing pipeline, or just better compression at build time. The comparison table above tells you which tool fits each answer. Run the numbers before you commit to a migration.