With so many S3-compatible providers on the market, it can be difficult to understand which one truly fits your workload.
To keep things objective, we used MinIO warp v1.0.7, a widely adopted open-source benchmarking tool. Tests included upload, download, mixed workloads, small object throughput, and large object handling. All benchmarks were run under identical conditions on a Debian 13 VM with 16 GB RAM, located in US-East-1, using 8 concurrent threads and consistent object sizes.
Our methodology is fully reproducible — every parameter and setup detail is publicly available. You can find the complete benchmark methodology and replication guide with step-by-step instructions to run your own tests. All pricing information reflects rates publicly available as of October 1, 2025.
Here's the summary:
A few things that surprised us:
- Mixed operations (simultaneous reads, writes, metadata) is the most realistic measure of real-world performance — and results vary wildly
- Cloudflare R2 has zero egress fees but came in last on mixed ops (44 Mbit/s) and small objects (42 obj/s)
- Hetzner's numbers look low — but that's because we tested from US-East-1 and they're EU-only, so transatlantic latency skews the results
- Backblaze B2 dominates on downloads but showed timeouts during large-object tests
For full per-provider breakdowns and detailed analysis: https://rabata.io/s3-comparison
Top comments (0)