Cloudflare's October 23 post makes Origin Range Requests a cache rule on all plans. The rulesets API note is September 2. The docs page was updated September 14.
On October 23 Cloudflare made Origin Range Requests a self-service cache rule on all plans. The API note is September 2. The figures are Cloudflare's.
The short version
Origin Range Requests, as a cache rule a customer turns on, on all plans, without a support ticket. Cloudflare says the old behavior could pull 128 MiB for a small seek, and that for R2 the new path cut p99 time to first byte by about half. A September 2 changelog already documented the API. This page does not print a request body. Two of the three sources are Cloudflare.
October 23 is the self-service post
Matthew Gumport's Cloudflare blog post on October 23, 2026 says Origin Range Requests are now a self-service feature on all plans. A customer turns the behavior on with a cache rule. There is no range size to tune and no support ticket, the post says. The problem it describes is an old cache path on which a player asking for a few megabytes in the middle of a file could cause Cloudflare to pull 128 MiB from the origin. The new path, the post says, fetches on the cache's own chunk boundaries and returns only the bytes the client asked for. That is the product claim. It is Cloudflare's.Data Agent Kit's date is a Google post, not this cache rule.
The numbers are the company's
The post says that for R2, p99 time to first byte dropped by about half, and that a follow-up cut tiered-cache traffic by about 6 percent for R2 and similar workloads. This page did not rerun those traces. A half and a 6 percent are not a promise for a video library that does not look like R2. The post also says a request smaller than the chunk can still fetch the whole chunk. The waste is bounded by the chunk, which is the point of the change, and it is not zero.the October 9 Workers profile is an earlier Cloudflare post.
September 2 and September 14 come first
The changelog entry dated September 2, 2026 says the rulesets API can set Origin Range Requests on a cache rule, with modes for on, off, and default. The cache docs page says it was last updated September 14, 2026, and it already describes origin range fetching aligned to the cache. So October 23 is not the first time the feature existed in public docs. It is the day Cloudflare told customers they can turn it on themselves on every plan. A reader who files the blog as the invention will miss a six-week API note.K-ISMS is a certificate program, not a cache rule.
What the docs say the feature will not do
The docs say the partial-file path is for a GET that is eligible for caching when it arrives. If the origin response turns out uncacheable, Cloudflare can still fetch a range and will not store the bytes. HEAD and other methods are not that path. An unsatisfiable range draws a 416. A range Cloudflare cannot safely compose is served as the whole representation. The feature does not make an ineligible response cacheable, and the maximum cacheable size is checked against the complete object, not against the small slice a client named. Those limits are the part a performance headline leaves out.a TLS certificate is a different binding.
The origin still has to cooperate
Cloudflare says an origin that ignores Range and returns a full 200 can be used, but the whole file has to come down. The changelog says origins should honor an identity content encoding and return consistent partial responses. That is a requirement on the origin, not a switch in the dashboard. A customer who enables the rule in front of an origin that does not do ranges has turned on a path that falls back to the full object. The post is honest about that cost. This page will not paste the JSON that sets the mode.
Who is in scope
Any plan, on Cloudflare's account, for cacheable objects that clients fetch in pieces: video, large downloads, package files, disk images. A site that never sends Range is not the audience. A site whose origin cannot answer ranges gets the fallback, not the saving. The 128 MiB story is the old failure mode the post wants readers to recognize, not a size every seek used to cost.
Two of three sources are the company
The blog, the docs, and the changelog are all Cloudflare. This page looked for a newsroom that measured the half and the 6 percent and did not file one. Summaries that reprint the post are not a second measurement. The same pattern has shown up here before, when a Cloudflare post was the record and an outside story was only a retelling. Here the outside retelling is not even in the sources. The record is the company's, and the dates inside it disagree with a launch-day reading of October 23.Node's October release is a different patch.
What a chunk means, without the body
The post says the cache already stores files in fixed pieces, and the range size became that piece instead of a customer setting. It gives one illustration, a request for a short span causing a fetch of the aligned piece, with the client still receiving only the span. This page does not turn that illustration into an API call. The impact is predictable extra work instead of a customer-chosen range that could pull far more than the viewer needed.
The limit
No request body. No claim of an independent benchmark. No claim that October 23 invented the feature. The record is a self-service announcement on top of a September API note, with performance numbers that belong to Cloudflare. The short answer this page will stand on is this: Origin Range Requests, as a cache rule a customer turns on, on all plans, without a support ticket. Cloudflare says the old behavior could pull 128 MiB for a small seek, and that for R2 the new path cut p99 time to first byte by about half. A September 2 changelog already documented the API. This page does not print a request body. Two of the three sources are Cloudflare.
Sources
The reports this brief is filing. Futureweb did not republish them.
Questions
Is October 23 the invention of Origin Range Requests?
No. The September 2 changelog already documents the rulesets API. October 23 is the post that says the feature is self-service on all plans.
What performance numbers does Cloudflare give?
For R2, p99 time to first byte about half of the old path, and a follow-up that cut tiered-cache traffic about 6 percent for R2 and similar workloads. Those are Cloudflare's measurements.
Does the feature make an uncacheable response cacheable?
Cloudflare says no. The origin still has to answer byte ranges. A small range does not bypass the maximum cacheable size.
Why are all three sources Cloudflare?
This page did not find a second newsroom that measured the feature. The blog, the docs, and the changelog are the company's own record, and the brief says so.
Does this page include the API example?
No. The changelog shows a JSON body. The news is that a cache rule can set the mode, not the body itself.