Antfly Circus Maximus

Benchmark views

Latest collects the newest reviewed result for every lane. Release pages keep evidence pinned to one Antfly version.

01 Antfly Circus — Latest The newest reviewed result available for each Antfly benchmark lane. Sections may use different Antfly releases or main commits; every evidence block names its exact Antfly subject. 11 complete · 7 partial · 0 valid negative · 7 missing 21 Aug 2026 — 8 Oct 2026 02 Antfly Circus — v0.2.1 wow such release. many ant. wow 7 complete · 1 partial · 0 valid negative · 0 missing 9 Sep 2026 — 9 Sep 2026 03 Antfly Circus — v0.2.1-rc.5 Five measured lanes for the official v0.2.1-rc.5 release (43fda0ba), on Linux ARM64 / rootless Podman, Apple M4 Max, 10 VM CPUs and 24 GiB. Peer numbers are historical references, not matched rc.5 measurements. MIRACL and STaRK were not rerun. 5 complete · 0 partial · 0 valid negative · 0 missing 8 Sep 2026 — 8 Sep 2026 04 Antfly Circus — v0.2.1-rc2 Fresh v0.2.1-rc2 honed evidence measured from the product-equivalent b4a46545 tree; the final tag adds only release-workflow fixes at 4be5a0e9. 6 complete · 2 partial · 0 valid negative · 0 missing 30 Aug 2026 — 3 Sep 2026 05 Antfly Circus — v0.2.1-rc1 Fresh Antfly-only standard honed evidence for v0.2.1-rc1 with reviewed competitor values and exact version provenance. 4 complete · 3 partial · 0 valid negative · 0 missing 27 Aug 2026 — 27 Aug 2026 06 Antfly Circus — v0.2.1-rc0 Exact-subject v0.2.1-rc0 evidence from the local Apple Silicon campaign, plus fresh Linux x86_64 rc0 Quickstart trials. The release-candidate phase ledger now appears in the page introduction rather than as a benchmark lane. 6 complete · 2 partial · 4 valid negative · 0 missing 25 Aug 2026 — 26 Aug 2026 07 Antfly Circus — v0.2.0 Benchmark results for Antfly v0.2.0. Measured lanes retain their own machine, completion time, runtime, and protocol; lanes without selected evidence remain visible. 7 complete · 1 partial · 0 valid negative · 13 missing 13 Aug 2026 — 21 Aug 2026 08 Antfly Circus — v0.1.3 Antfly-only benchmark evidence for the published v0.1.3 Linux ARM64 release, including three fresh VectorDBBench lifecycles, one completed CRAG footprint cell, and a graph correctness-gate outcome. 1 complete · 1 partial · 1 valid negative · 0 missing 18 Aug 2026 — 18 Aug 2026 40 Parse Atlas Reviewed page-level evidence for Antfly PDF parsing, rendering, OCR, and extraction failures. 2 reviewed runs · newest: pdf-hard 12 page cards · 94604a4684 50 Circus Cinema Interactive terminal recordings from product quickstart runs. 16 recordings · benchmark-tagged Protocols 3–4 · Pi 0.78.0
Coverage Every number traces to its run

Each published result links back to the recorded run that produced it — the Antfly build, the protocol, and the retained artifacts. Missing lanes explain what evidence is unavailable.

Provenance Per-section run stamp

Cross-date and cross-machine evidence stays visible rather than being flattened into a false matrix.

Payload Reviewed projections

Tables and charts contain allowlisted metrics. Cinema recordings explicitly publish reviewed terminal and session logs; other raw rows, prompts, documents, and machine-local paths stay private.