apply for beta
menu
benchmark library

ripgrep benchmark

historical measurement · 2026-10-05

One cold run on Sorus and one on GitHub-hosted infrastructure. Both completed successfully. Hardware and measurement limitations are listed below.

Historical result. The machines have different vCPU counts. One run per platform is insufficient to establish average or current-release performance.

recorded duration

Durations use the historical command-to-outcome method. They exclude parts of the time-to-result measured by the newer harness. No price comparison is available.

workload details

public source
BurntSushi/ripgrep
commit
3fce3b5bb0236da2df6d99672afb8a719642eca7
Cargo invocation
cargo test --workspace --locked
toolchain
1.98.0
cache state
No restored Cargo home, target directory or build cache. Baseline: No cache restore step, rust-cache or sccache.
sample count
1 per side · no distribution available
release qualification
Historical run; not qualified as the current beta release.

measurement limitations

  • One paired observation (n = 1 per side). No percentiles, repeatability estimate or warm-state measurement.
  • Historical command-to-outcome boundary: the Sorus client invocation and hosted command steps are not the newer full time-to-result boundary.
  • Unequal sizes: Sorus 8 vCPU versus GitHub-hosted 2 vCPU. Do not interpret this as an equal-hardware speedup.
  • Different host CPUs and environments. Matching a vCPU count does not match per-core performance.
  • One Sorus host; no fleet, saturation or concurrency claim.
  • The underlying evidence bundle is retained internally. This download is an allowlisted summary, not the complete original raw evidence.

download measurement data

The JSON contains the timing fields and context shown on this page, plus the original record digest. It excludes internal operational identifiers and cost-model figures.

download record JSON

The full raw execution bundle is not yet available for public download.

read the method