CI test results
How is ZeroWeb doing?
Scheduled CI runs the same compatibility and performance checks over time. This page reads the results committed to the repository.
Results updated Nov 22, 2026
Full WPT accounting
Full WPT pass rate
The denominator is the entire upstream WPT corpus — the union of the three mainline browsers’ aligned stable runs. Suites ZeroWeb has not imported yet count as zero, so the uncovered subtests below are the actual roadmap gap, directory by directory.
Biggest gaps by directory
css91,075 uncovered subtestshtml53,730 uncovered subtestswebnn14,210 uncovered subtestssvg14,152 uncovered subtestsfetch10,967 uncovered subtests
All directory gaps
Scroll horizontally to see every column, including dates and evidence.
| Directory | Subtests | Passed | Uncovered | Env-limited | Files |
|---|---|---|---|---|---|
| css | 199,069 | 107,994 | 91,075 | 0 | 6,970 / 36,176 |
| html | 160,124 | 106,394 | 53,730 | 0 | 6,395 / 10,979 |
| webnn | 14,210 | 0 | 14,210 | 0 | 0 / 610 |
| svg | 15,510 | 1,358 | 14,152 | 0 | 486 / 1,546 |
| fetch | 13,336 | 2,369 | 10,967 | 0 | 152 / 963 |
| webdriver | 10,520 | 0 | 10,520 | 0 | 0 / 691 |
| dom | 67,148 | 57,300 | 9,848 | 0 | 359 / 863 |
| referrer-policy | 8,815 | 1 | 8,814 | 0 | 37 / 1,483 |
| url | 12,443 | 4,727 | 7,716 | 0 | 36 / 79 |
| editing | 117,246 | 110,034 | 7,212 | 0 | 692 / 889 |
| quirks | 6,984 | 40 | 6,944 | 0 | 6 / 55 |
| shadow-dom | 15,368 | 8,518 | 6,850 | 0 | 172 / 387 |
| streams | 5,796 | 912 | 4,884 | 0 | 69 / 318 |
| IndexedDB | 5,853 | 1,072 | 4,781 | 0 | 168 / 820 |
| webaudio | 5,892 | 1,402 | 4,490 | 0 | 50 / 356 |
| content-security-policy | 4,565 | 199 | 4,366 | 0 | 447 / 957 |
| mathml | 4,321 | 0 | 4,321 | 0 | 0 / 626 |
| webcodecs | 3,972 | 0 | 3,972 | 0 | 0 / 322 |
| sanitizer-api | 3,834 | 0 | 3,834 | 0 | 0 / 51 |
| trusted-types | 3,764 | 0 | 3,764 | 0 | 0 / 249 |
| web-animations | 4,280 | 525 | 3,755 | 0 | 67 / 208 |
| domparsing | 3,757 | 121 | 3,636 | 0 | 25 / 100 |
| urlpattern | 3,266 | 0 | 3,266 | 0 | 0 / 30 |
| service-workers | 3,277 | 463 | 2,814 | 0 | 107 / 336 |
| mimesniff | 3,886 | 1,296 | 2,590 | 0 | 1 / 7 |
| webrtc | 2,424 | 0 | 2,424 | 0 | 0 / 270 |
| mixed-content | 2,281 | 0 | 2,281 | 0 | 2 / 388 |
| WebCryptoAPI | 70,924 | 68,677 | 2,247 | 0 | 326 / 340 |
| webgl | 4,289 | 2,112 | 2,177 | 0 | 9 / 28 |
| scroll-animations | 2,167 | 0 | 2,167 | 0 | 0 / 291 |
Mainline browsers, own full runs
Scroll horizontally to see every column, including dates and evidence.
| Suite | Pass rate | Subtests |
|---|---|---|
| chrome | 97.25% | 2,211,408 / 2,273,881 |
| firefox | 95.47% | 2,148,124 / 2,249,945 |
| safari | 94% | 2,122,523 / 2,258,045 |
| ladybird | 96.5% | 2,172,954 / 2,251,800 |
Rendering self-consistency
Self-source reftest trend
Both test and reference pages are rendered by ZeroWeb. Each point is the latest result for that month; this measures self-consistency. Corpus versions and passed / total counts are in the table.
Monthly measurements
Scroll horizontally to see every column, including dates and evidence.
| Date | Pass rate | Self-consistency reftest | Corpus |
|---|---|---|---|
| Aug 2026 | 82.9% | 13,941 / 16,814 | v1.10 |
| Sep 2026 | 91.8% | 14,748 / 16,058 | v1.10 |
| Oct 2026 | 95.4% | 15,323 / 16,058 | v1.10 |
| Nov 2026 | 95.4% | 15,316 / 16,058 | v1.10 |
Coverage by WPT suite
All tracked suites
One row per WPT suite. Denominators differ across rows (reftest files vs. testharness subtests) and the unit is shown next to every number. Rows marked “Planned” have no data yet and will fill in as those goals run.
Suite measurements and evidence
Scroll horizontally to see every column, including dates and evidence.
| Suite | WPT scope | Passed / total | Pass rate | Status | Date | Evidence |
|---|---|---|---|---|---|---|
| XHTML syntax (entity matrix)wptfyi-html DC-1 trial: new testharness-xhtml-syntax runner channel; 12 entity-matrix outer pages 25606/25606 all pass with per-file counts matching the aligned browser denominator digit-for-digit (11 x application/xhtml+xml files at 2125 each + text/html file at 2231); external-entity boundary verified offline (bogus systemId foo; zero DTD fetches; built-in entity tables only). Mechanism note: DOMParser XHTML face resolves via HTML parse fallback (built-in entity superset; assertion face indistinguishable); true .xhtml document load path (xml-stylesheet 1 subtest + 2 reftest faces) parked for DC-2. Zero-execution accounting: support/ assets + rel=match reftests + .xhtml. Wired to weekly CI via catch-all-8. | html/the-xhtml-syntax | 25,606 / 25,606 subtests | 100% | Active | Nov 20, 2026 | GitHub ↗ |
| WebAssembly core (spec testsuite)M1 first full baseline on pinned 315976933870 (258 wrapper pages; 273 cases incl webapi). 99.98% overall - MVP/simd/gc/exceptions/relaxed-simd/bulk-memory/multi-memory all green on native V8 via restored page env; async wasm compile quartet rewired over sync primitives at runner warmup (platform-pump stall root cause; jsapi 4 booked timeouts cleared 715+4T -> 878/878 zero regression). Residual 9F memory64 64-bit index face. | wasm/core | 188,280 / 188,289 subtests | 100% | Active | Oct 29, 2026 | GitHub ↗ |
| WebAssembly (jsapi floor re-baseline)Floor raise booked under wasm-compat: the 2026-09-12 715/719 baseline carried 4 timeouts whose root cause (native async wasm compile never completing under the runner page loop) is fixed at the runner warmup restore point; channel now 878/878 exit 0, all 715 baseline subtests preserved byte-for-byte. | wasm/jsapi | 878 / 878 subtests | 100% | Active | Oct 29, 2026 | GitHub ↗ |
| CSS rendering compatibilityR5603 accounting (11-15): corpus 17,049 at 995F with FIXED=1 REGRESSED=0 — the fail rise 691→995 tracks the denominator growth 16,594→17,049 (anchor-position corpus batches), not regressions; the flex-item va real regression from the g50 knife was bisected to 7bc31687d and fixed in-round; passes 16,054 = 17,049 − 995 (denominator minus fails). | css2 / CSS2 / flexbox / grid / position / tables / multicol / text / writing-modes / fonts / text-decor | 16,054 / 17,049 tests | 94.2% | Active | Nov 15, 2026 | GitHub ↗ |
| SVG document modelM4 closeout: SMIL ledger finalized at post-S3 baseline - animations Pass 4->48 subtests (scaffold release, zero is-not-a-function dead face), 228 Fail subtests bucketed: 199 charged to SMIL clock (goal-excluded face: timing/value 155 + path d interpolation 33 + clock grammar 7 + interval misc 4), 22 getCTM carriers to rendering-compat, 7 SVGLength context units to S2a FIXME batch, 7 color-interpolation to rendering/style; 32 Timeouts bucketed (seeking 8/eventbase 6/interval 7/syncbase 5/sampling 3/rendering-interaction 3, 29 charged to SMIL clock). Collected: SVGAnimationElement getStartTime/getSimpleDuration exception semantics (no active interval -> InvalidStateError, no dur -> NotSupportedError; target exceptions case 2/2 green from baseline Fail). Full corpus 1238/1767 = 70.1% -> 1245/1772 = 70.3% (+2 Fail->Pass, zero Pass regression zero DOWN); probe-excluded same-denominator 69.0%->69.1%. DC-4 four gates all green (fmt + clippy -D warnings + reftest 712/712 + make test 76 suites). Goal closed: M1-M4 complete, seven-item booked list finalized. | svg | 1,245 / 1,772 subtests | 70.3% | Active | Oct 22, 2026 | GitHub ↗ |
| Selection API (expanded corpus)Phase-2 M1 corpus expansion: selection/ recursive + editing dirs -> 675 cases / 51,757 subtests (68.87%); phase-1 anchor preserved (25 cases match wpt.fyi 3-browser subtest counts case-by-case; 2968/2970 face intact = net-zero regression). First-run deleteFromDocument-HTMLDetails flip root-caused in-goal (a8ec673b1 spec-contradicting deleteContents ec tail-boundary whole-delete rule, exposed by engine text-view fixes enabling the editing-host marker-walk climb) and fixed in-place same round; canonical 35,638P/16,115F/4N, selection face 22,304P/13,755F clustered into 8 mutually-exclusive buckets closing on the full face (C1 boundary-validation 13,277F -> D2 proposal; C2 modify() 190F -> M2 re-judge; C8 -> 0F fixed). Coverage re-test: 158/194 universe files covered, plain/query-variant uncovered = 0 (idlharness 112 subtests ledger-excluded, crashtests 35 zero-weight). fontSize x3 flip attributed to editing goal (90c992231) and booked for transfer. | selection / editing | 35,638 / 51,757 subtests | 68.9% | Active | Nov 4, 2026 | GitHub ↗ |
| HTML semantics (forms/tabular/etc.)M2 slice 1 (input value IDL face): full nine-domain corpus re-run 3248/6609 (+533 vs M1; forms 1897->2320; the-input-element face 528->951); zero regressions vs baseline per-subtest diff across all nine domains; 81 files improved. Total +240 subtests = more files executing this round (maxlength.html etc.). +15 Timeout/+14 NotRun in already-failing files attributed to concurrent bench measurement window (regression-neutral). | html/semantics | 3,248 / 6,609 subtests | 49.1% | Active | Oct 15, 2026 | GitHub ↗ |
| CSS testharness (CSSOM/style API)M4/DC-4 closeout: CI coverage run 41005658107 css-face-a/b/c shards all green; 16/16 dir case counts match local baseline; CI subtests 65446/17759 vs corrected local 65501/17781 (-0.08%/-0.12% timing band; no helper collapse after css/support whitelist fix e425f845a). css-sizing denominator re-measured 2238 (fresh local == CI per-case exact; batch-1 row 1894 was single-process run-order residue, superseded by this row). Coverage snapshot (locally aggregated from 17 shard artifacts; CI aggregate skipped by form-encode shard failure in the encoding domain): css topdir 3293 files / 17750 passes / 9.08%, up from 61/232/0.12% at DC-3; pilot dirs zw_passes > 0. | css-values / selectors / css-variables / css-color / css-align / css-sizing / css-typed-om / css-fonts / css-shapes / css-anchor-position / css-masking / css-backgrounds / css-transforms / motion / css-grid / css-text | 17,781 / 65,501 subtests | 27.1% | Active | Oct 30, 2026 | GitHub ↗ |
| Canvas 2D919 upstream files imported; 832 executable testharness tests all pass. Chromium oracle strict true-pass sample 34/41 (82.9%). | html/canvas | 832 / 832 tests | 100% | Completed | Aug 16, 2026 | GitHub ↗ |
| Console IDL & formattingGoal closed: DC-1/2/3 all met. Corpus 28/28 = 100% held; make test 71 suites green; make reftest 710/710 (100%); product-smoke welcome oracle 15.35% (< 20 gate) + struct-check PASS all viewports. | console | 28 / 28 subtests | 100% | Completed | Oct 14, 2026 | GitHub ↗ |
| Encoding (TextEncoder/Decoder)Independent revalidation on combined main b2ce0f883 (7 render-stream commits atop round-2 fix 825549020, zero encoding-face source change): 12072/12075 = 99.98% byte-identical to round-2 close; the 3 fails match booked residuals exactly (textdecoder-copy SAB x1, streams/decode-utf8 transferred chunk x2) - zero regression. Note: make target exits 1 by design when residuals exist (not a land gate); read subtest counts, not exit code. | encoding | 12,072 / 12,075 subtests | 100% | Completed | Oct 19, 2026 | GitHub ↗ |
| Legacy multibyte document decodingDC-4 coverage rerun (local full run: 51 Makefile channels + CI-parity canvas/canvas-worker/constraints): decode face 360431/360431 = 100% reconfirmed; wpt.fyi four-family denominator 1128104 counted in coverage-2026-10-29.md - decode-face uncovered cleared; remainder 767673 = encode face adjudicated not-run per M3 (form/URL-compat gaps). Full-corpus headline 7.41% -> 24.19%. | encoding/legacy-mb-japanese / encoding/legacy-mb-korean / encoding/legacy-mb-schinese / encoding/legacy-mb-tchinese | 360,431 / 360,431 subtests | 100% | Completed | Oct 29, 2026 | GitHub ↗ |
| Form & URL legacy encoding serializationM2+M3 closeout: form face (acceptCharset serialization: encoding override + NCR fallback + USV conversion) + href face (encoding-parse pre-encode at a/area getters + host-bridge NCR direct-out %26%23N%3B + query-set % normalization) -> 766958/766959 = 99.9999% from 0P baseline; sole non-pass = data:iframe runner Timeout (contract hold). DC-4 coverage rerun: full-corpus headline 24.19% -> 58.02%; encoding topdir 31.38% -> 97.93%; delta 766958 exact vs M2 baseline; decode face 360431 byte-identical zero regression. | encoding/legacy-mb-japanese / encoding/legacy-mb-korean / encoding/legacy-mb-schinese / encoding/legacy-mb-tchinese | 766,958 / 766,959 subtests | 100% | Completed | Oct 29, 2026 | GitHub ↗ |
| Form validation45 constraints cases + requestSubmit + checkValidity at v1 close. | html/semantics/forms | 919 / 919 subtests | 100% | Completed | Aug 17, 2026 | GitHub ↗ |
| DOM (js-dom)11 failures + 16 timeouts counted in denominator (R388 full sweep). | dom | 55,808 / 55,835 subtests | 100% | Completed | Aug 31, 2026 | GitHub ↗ |
| Web Audio50 test files / 1418 subtests all pass. | webaudio | 1,418 / 1,418 subtests | 100% | Completed | Sep 5, 2026 | GitHub ↗ |
| Beacon/Visibility/onlineGoal closed (DC-1~3 met): 41/41 subtests = 100%; post-archive verification round 2026-10-18 re-ran all gates on main: make test 21814P/0F (T58 fmt gate drift on beacon fixture fixed as pure-formatting style commit) + reftest 712/712 + product-smoke welcome 17.07% (81944/480000 px) double-run byte-constant; welcome-800 PNG sha256 c483bd9c95f82793 identical to closure round. Deviations booked: FIXME(beacon-mixed-content); FIXME(keepalive-quota-inflight). Archived under docs/goal/archive/page-state-compat/. | beacon / page-visibility / online | 41 / 41 subtests | 100% | Completed | Oct 18, 2026 | GitHub ↗ |
| Pointer LockConstructor residual slice: Event interface eventPhase constants (NONE=0/CAPTURING_PHASE=1/AT_TARGET=2/BUBBLING_PHASE=3) installed on native V8 bindings (dom_bindings/event.rs attach_event_constants, interface object statics + Event.prototype, WebIDL es-constants) - page realms' globalThis.Event is overwritten by register_ctor so shim part05 R23 constants never took effect in production pages (R384 native-only path). Channel 21/22 -> 22/22 (100%), files 18/18 first all-green. Side gain: dom channel Event face +21 subtests green (251 -> 230 fails A/B same-filter, zero new red: Event-constants x4 / bubbles x10 / dispatch order+target x4 / EventListenerOptions-capture x2 - all assert or filter on eventPhase constants); uievents wheel family 3/3 zero regression; make test green; reftest 712/712. Ledger correction: file counts from Commit 4 onward undercounted by 1 (wheel x2 are 2 subtests of ONE file counted as two files) - Commit 5 was actually 16/18, TBD-3 17/18; prior rows/evidence kept as recorded, see evidence section 4. | pointerlock | 22 / 22 subtests | 100% | Completed | Oct 19, 2026 | GitHub ↗ |
| Service Workers (subset)121 cases across core/fetch/CacheStorage runners; deterministic double-run green. | service-workers | 652 / 652 subtests | 100% | Completed | Sep 6, 2026 | GitHub ↗ |
| Cache API39 test files / 449 subtests (window scope). | cache-storage | 449 / 449 subtests | 100% | Completed | Sep 6, 2026 | GitHub ↗ |
| IndexedDB168 of 210 upstream files imported at pinned revision; zero skips. | IndexedDB | 1,073 / 1,073 subtests | 100% | Completed | Aug 19, 2026 | GitHub ↗ |
| WebAssembly WebAPI (streaming convergence)S18 handoff clearance: residual 2F (ReadableStream body with bare ArrayBuffer chunk must TypeError) settled by Body mixin arrayBuffer() returning spec ArrayBuffer - enqueued product hits the existing body-consume chunk type gate, no new gate. Supersedes the M3 row (270/272 = 99.26%). jsapi floor 878/878 intact; memory64 9F unchanged (Chrome-identical declaration-limit account). | wasm/webapi | 272 / 272 subtests | 100% | Completed | Oct 29, 2026 | GitHub ↗ |
| Web Storage (localStorage/sessionStorage)Goal close at M3 verdict (DC-1~4 all satisfied): 1251/1251 = 100% on the 25-case window-executable subset (trajectory 94.9% -> 97.6% -> 99.0% -> 99.4% -> 100%; monotonic no down-round). M3 StorageEvent interface fidelity: constructor no-Call TypeError + required type + USVString url null->'null' + initStorageEvent 8-arg reset semantics. Excluded faces fully classified (30 cases: event dispatch family 19 + window.open 4 + cross-origin iframe 1 + partition/document-domain 4 + manual 2) charged to navigation-compat M3 frame-tree gate — 100% is the executable-subset scope honestly noted. Gates: make test decomposed legs all 0 failed (browser xvfb leg silent-kill attributed to parallel-stream contention via isolated rerun LEG-EXIT=0) + make reftest 708/708 zero regression + product-smoke welcome oracle diff 15.40% <= 20% + fmt/clippy clean. | webstorage | 1,251 / 1,251 subtests | 100% | Completed | Oct 11, 2026 | GitHub ↗ |
| HTML document face (syntax/DOM)R5331 cross-domain check: pre/post revert comparison on f488fc8cf is bit-exact (61341P/62F/7T/1U both sides) - R5331 zero impact on this channel. Current-pin account supersedes the 2026-10-20 closeout (61347/61422 at older pin); delta (html5lib 3T->3F, render-blocking 58F+10T->55F+7T, denominator -11) stems from pin roll and interim engine land by other streams, not R5331. | html/syntax / html/dom | 61,341 / 61,411 subtests | 99.9% | Completed | Oct 30, 2026 | GitHub ↗ |
| WebGL (conformance)M4 closeout signed off and archived (user sign-off 2026-10-23: WebGL2/WebGPU deferral with reopen conditions, CPU-path accounting retained, 3 upstream IDL-mixin fails accepted as baseline). Final baseline 2112/2115; reftest 716/716 incl. clear/draw scale twins; make test exit 0, clippy/fmt clean. | webgl | 2,112 / 2,115 subtests | 99.9% | Completed | Oct 23, 2026 | GitHub ↗ |
| Editing & selection5 failures counted in denominator; baseline at goal start was 45 pass (1.7%). | selection | 2,994 / 2,999 subtests | 99.8% | Completed | Sep 8, 2026 | GitHub ↗ |
| WebAssembly (jsapi)Channel restoration via encoding-compat round-2 revalidation: b785548dc (2026-10-17) switched runner warmup to execute_script_with_dom whose dom polyfill REPLACES native V8 WebAssembly with the host auto-bridge (production execute-path contract) - the bridge stub Module/exports collapsed the channel to 141/460 (whole-face fails: Branding/name/length/Empty buffer). The 715/719 baseline (2026-09-12) ran on the native V8 interface (pre-polyfill warmup). Fix without touching the production bridge: bridge install stashes the pre-clobber native reference under a non-enumerable __zwNative__ key; runner warmup restores native wholesale after with_dom (exact pre-10-17 page-environment shape for all corpora). Verified 715 Pass / 0 Fail / 4 Timeout - byte-identical to baseline incl. the 4 booked timeouts; encoding corpus 12072/12075 zero regression. Production __WASM_BRIDGE__ semantics unchanged; resident unit test locks the stash+restore key pair. | wasm/jsapi | 715 / 719 subtests | 99.4% | Completed | Oct 19, 2026 | GitHub ↗ |
| DOM expansion (Range API + html/dom interface surface)Goal closure (DC-1..4 all satisfied): ranges top-level semantic accuracy 99.07%->99.81% (M2 dataChange 369F->0F + M3 mutations tail 8F->0F via R262/R263 registry-index completeness guard; 4T CASE_TIMEOUT truncation parked); html/dom old face zero regression (61F parked composition unchanged; head/wip A/B subtest sets identical) and new face 469P/636F->621P/484F via documents 136F->P + dataset 16F->P; residual 484F fully attributed (innerText 400F layout-dependent parked + partial-updates 56F proposal + dir 13F + documents 3F runner-pipeline + misc 12F) plus ranges tentative 189F interface-missing parked. | dom/ranges / html/dom | 100,092 / 100,831 subtests | 99.3% | Completed | Nov 5, 2026 | GitHub ↗ |
| Encoding phase 2 (root-face importables + coverage closeout)Goal-37 closeout at M4/DC-4: 1139476/1152394 = 98.879% (gate 98.8%); passes exactly equal the booked ceiling (denominator minus 12889 structural-hold pool minus 29 pin-drift/API/contract residuals); 14 non-pass all booked residuals (3F SAB + 10F iso-2022-jp-encoder streaming-model cluster + 1T data:iframe). Form-face 550054 pool was blocked by an engine-shared-path defect (shim pending-ghost double render + downstream named-target submit no-op); unblocked by sibling-stream s168 shim fix (daf3bc4c5 R383 cross-generation bucket skip + R383a paired-env latch + webview cursor publish) with zero channel/engine changes from this goal; 58-job batch all rc=0 in one pass. Zero regression: 68-file encode face 766958P/1T byte-identical to goal-36 ledger, decode 360431P/0F, enc30 12072P/3F. | encoding | 1,139,476 / 1,152,394 subtests | 98.9% | Completed | Oct 31, 2026 | GitHub ↗ |
| File API (Blob/File/FileReader/objectURL)Goal closure at M3 (DC-1~3 verified item by item): 493/499 = 98.8%, rerun constant with M2-P5 (two-consecutive-rounds-no-decline gate met). Trajectory 311/499 (62.3%, M1) -> 435 (87.2%, M2-P1) -> 467 (93.6%, M2-P3) -> 488 (97.8%, M2-P4) -> 493 (98.8%, M2-P5 + closure rerun). Residual 6F all cross-domain booked: url-with-fetch/xhr Request/revoke lifetime 5 (net-api) + FrozenArray<MessagePort> e.ports 1 (webmessaging). Gates: make test 71 suites green + make reftest 708/708 zero inconsistent + product-smoke 15.40% <= 20% with struct PASS + product-smoke-legacy 39/39 struct PASS. | FileAPI | 493 / 499 subtests | 98.8% | Completed | Oct 11, 2026 | GitHub ↗ |
| Web Messaging (postMessage/Channel)Goal closure at M3 (DC-1~3 verified item by item): executable face 133/135 = 98.5%, trajectory 87/134 (64.9%, M1) -> 126/135 (93.3%, M2-P1) -> 133/135 (98.5%, M2-P2+P3); failure clusters zeroed, residual 2 non-pass parked (garbage-collected GC-close Timeout, user-activation missing navigator.userActivation). Cross-document family re-entry gated on navigation-compat M3 frame tree; worker-side messaging belongs to workers-compat. Gates: make test 71 suites 20539P/0F green + make reftest 708/708 zero inconsistent + product-smoke 15.40% <= 20% struct PASS x3 (same value as fileapi-compat closure) + product-smoke-legacy 42 fixtures 0 struct failures. testharness-dom double-run 54303/54304P of 54492, diff exactly one Event-timestamp flake family member - zero regression. | webmessaging | 133 / 135 subtests | 98.5% | Completed | Oct 12, 2026 | GitHub ↗ |
| Drag and DropM3 closure at pin 31597693: 44 cases / 446 subtests 437P/4F/5T (98.0%), 213P (55.0%) -> 437P (+224) across M2 convergence (+174) and M3 pipeline slices s1-s5 (+25: UA drag-gesture barrier x24, boundary relatedTarget + initial target update x1). Residuals booked to owning domains: 3 input drag-select (text-selection domain, classified in selection-compat), 1 shadow-DOM reparent (web-components/rendering coordination), 2 input_image injection trace timeouts + 1 pointer-suppression full face (runner infra), 2 js-file DataTransfer File face (out of M3 scope). | html/editing/dnd | 437 / 446 subtests | 98% | Completed | Oct 18, 2026 | GitHub ↗ |
| Media elements13 partial-pass counted in denominator; 209 test files. | html/semantics/embedded-content/media-elements | 640 / 653 subtests | 98% | Completed | Sep 4, 2026 | GitHub ↗ |
| Web Crypto phase 3 (promotion-domain convergence + zero-legacy endgame)Goal-38 phase-3 zero-legacy closeout: D1 user ruling (2026-11-02, 12-14 crates all RustCrypto-official; X448 via stable x448 0.6.0, Ed448 descoped - MSRV 1.85 no signing impl; KMAC handwritten over official cshake) then five domain slices to completion (Argon2/AES-OCB/ChaCha20-Poly1305; XOF x12; PQC ML-KEM/ML-DSA; X448/KMAC; Hybrid PQ/T KEM x3 combos), all-account Timeouts cleared to 0; no-dep slices M2-1..5 (supports gate +543P, scattered 5 all green, worker legs +13059P, 1T retired 13P, raw format aliases +80P). Rebase terminal ledger at HEAD layout after retiring 50 old .tentative. files (98 cases/14728 keys): 262 cases / 69795 keys = 67845P + 1950F + 0T, bit-CLEAN vs d1s5 anchor cut to HEAD subset; continuity vs phase-1 m6-residual terminal anchor: stable 74 cases / 25768 shared keys bit-identical zero migration; DC-4 coverage re-test unmapped 108 to 10 (tentative dimension zeroed; residual 10 = sliced-variant channel-infrastructure account, universe-missed 82 all named); DC-5 lineage zero handover: all non-P = Ed448 descoped named (14 cases), no parked debt moved to any other goal; make test reds all attributed to sibling streams (engine lib 5 CSSOM serialization s249 drift + quickjs renderer lib 1 timing + browser bin forms/typing 1F, combined-tree collision ledger; goal face itself all green); rally-goal-38 launcher deleted at archive per 2026-11-01 user decision. | WebCryptoAPI | 67,845 / 69,795 subtests | 97.2% | Completed | Nov 3, 2026 | GitHub ↗ |
| Web Workers (dedicated)Goal close at M4 verdict (DC-1~4 satisfied item by item; workers channel rerun 693/714 exact reproduction on 175 cases; make test green + clippy -D warnings + fmt clean + make reftest 712/712 zero regression). Residual 21 non-pass (12F+8T+1PF) all classified non-blocking: P8 runner infra 6F+3T (302 redirect + synthetic origin) + URL encoding 2F + terminate depth 1F+1T (nested account) + SharedWorker exposure 1F (66-case exclusion account) + excluded-face accounting 4T (P6) + resource timing/nested sync XHR 2F + ReadableStream battery 1PF. Parking finalized: SharedWorker not implemented (re-entry on isolate/threading) + nested worker implemented (terminate depth inherent boundary) + module worker 41 cases excluded (module graph loader prerequisite). quickjs behavior matrix not evaluated (all behavior evidence v8 leg; quickjs leg clippy compile gate only) recorded. | workers / dedicated-workers | 693 / 714 subtests | 97.1% | Completed | Oct 21, 2026 | GitHub ↗ |
| Web Crypto full corpus CI integration (wptfyi 90-plan goal 5)wptfyi goal-5 closeout: upstream WebCryptoAPI full corpus wired into the weekly crypto CI shard (fetch-wpt-webcrypto, 38-corpus v3 lineage); coverage snapshot coverage-2026-11-09 (aligned stable 792c7512e3) records 67877/70920 = 95.71% subtests, net +41098 >= 31000 budget; DC-2 generateKey [EnforceRange] fix (+800) lands with the next weekly run; neighborhood zero attributable regression (shadow-dom wobble forensically attributed to page-timeout truncation). | WebCryptoAPI | 67,877 / 70,920 subtests | 95.7% | Completed | Nov 10, 2026 | GitHub ↗ |
| Origin Private File System13 WPT fs cases (window scope); 7 failures counted in denominator. In-memory shim baseline was 48/133 (36.1%). | fs | 126 / 133 subtests | 94.7% | Completed | Sep 9, 2026 | GitHub ↗ |
| Touch Events (constructors/attributes)Goal closed (DC-1~3 met): executable face 30P + 2 neutral PF = non-neutral 100%; make test 71 suites green + reftest 710/710 + product-smoke 15.35% (struct-check PASS). Injection face booked with numbers: 9 cases / 28 subtests 1P/23F/4T (single root cause - no host touch truth stream) with re-entry = host touch injection pipeline. | touch-events | 30 / 32 subtests | 93.8% | Completed | Oct 14, 2026 | GitHub ↗ |
| Editing commands (execCommand/queryCommand)DC-5 coverage retest closeout: both legs per-case --json runs through the gap pipeline (bin 40014d7d, judge 9 anchors zero drift vs e24). editing topdir 11.41% (2026-11-05 snapshot) -> 92.5% (+81.09pp); editing/run file coverage 0/120 -> 119/127 (98.49% subtests); editing/other 2/510 -> 462/529 (77.85%). Three-state clearance: executed 581 + named-excluded 47 (https 4 / iframe 43) + upstream-absent 260 (out-of-scope 232 / pin-missing 28) = universe 888, unaccounted 0. Uncovered 8796 = excluded/absent denominators 5425 + real reds 2878 (booked to wptfyi-editing stream d-rounds) + pin-drift 493. Denominator = wpt.fyi aligned stable union @71b58bb191 (revision drift vs snapshot acknowledged). | editing/run / editing/other | 108,448 / 117,244 subtests | 92.5% | Completed | Nov 19, 2026 | GitHub ↗ |
| Fetch & XHR & streamsPost-close S17 + S18 update; supersedes the 2026-10-29 terminal row (11011/12077). Master bookkeeping: M1 baseline 9354/12078 -> 11040/12077 = 91.4% (+1686 = S5-S16 +168 + S17 +28 real gains + S18 +1 baseline-drift correction; S13 denominator -1). S18 slice itself nets 0/0/0 - the +1 is the S17 stale-runner baseline correction adjudicated by triple experiment (S18 evidence bookkeeping 3). request-cache 98F attribution pending user decision, non-blocking. | fetch / xhr / url / mimesniff / streams / eventsource / websockets | 11,040 / 12,077 subtests | 91.4% | Completed | Oct 29, 2026 | GitHub ↗ |
| Shadow DOM corpus expansion (declarative materialization closeout)Goal 42 closeout: corpus roll + declarative shadow DOM materialization. Mixed measurement = full corpus x1 (9,856 subtests 8,437P, 365 files) with gethtml.html re-measured at x4 timeout multiplier (6,656P/0F, serial-matrix tail 252 truncated by collection budget, common subtests zero drift). shadow-dom topdir 4.05% (621P) -> 53.84% (8,264P/15,348, gate >=31.4% at 1.71x); custom-elements zero subtest regression (3,848 set-identical, 2,821P both sides, 64.76% vs ledger 64.67%); headline 8,210->8,437P no drop; attachment declarative materialization solo 654P/0F with full-corpus A/B P->F=0; adoptedStyleSheets residual 9 files 5P/23F routed to rendering-compat. | custom-elements / shadow-dom / the-template-element | 11,668 / 13,087 subtests | 89.2% | Completed | Nov 7, 2026 | GitHub ↗ |
| UI & pointer eventsGoal closed at tail cluster 41 (DC-1~4 met); 170 failures + 96 timeouts cataloged as parked debt in 14 categories with re-entry conditions. | uievents / pointerevents | 1,794 / 2,060 subtests | 87.1% | Completed | Oct 7, 2026 | GitHub ↗ |
| Clipboard & fullscreenclipboard-apis 61/73 (83.6%) + fullscreen 132/150 (88.0%) combined; DnD excluded pending host drag-input pipeline. | clipboard-apis / fullscreen | 193 / 223 subtests | 86.5% | Completed | Sep 24, 2026 | GitHub ↗ |
| Navigation & browsing contextsM3 phase close (gated minimal face approved 2026-10-21 '批准 M3 最小面'): S1~S7 seven slices - S1 srcdoc processing + sandbox directives (ASCII parse + allow-scripts gate) + servedDoc maturity dual-read + frames named access + child location write + load trust face; S2 doc.location==win.location identity + full URL parse face + child hash write; S3 joint iframe session history (lazy seeding + traversal routing, top ledger un-inflated); S4 replace-before-load minimal face ({{GET}} templates + child-script history param + frame-local back routing); S5 parent->child postMessage + child onmessage binding (with(window) conditional wrap); S6 child-window local history traversal (_zwDocBase document-generation model + async trusted popstate + out-of-bounds no-op) + parser-phase iframe autoload + runInlineScripts allow-scripts gate (also closed sandbox-ascii-case-insensitive); S7 nested iframe autoload in child docs (finish-pipeline recursion + per-document src resolution + depth cap 5 + sandbox flag inheritance) + lazy honored only in script-enabled documents + child-tree contentDocument/contentWindow face (_zwMEl lazy getters + cross-origin null gate). the-iframe-element 16P -> 37P; full corpus 437/476 (M2 final 91.8%) -> 473P/68F/43T = 584 case lines; per-path exact diff zero Pass regression across all seven slices. DC-1/2/3/4 all satisfied (DC-3 minimal-face reading per approval); gates make test 76 suites 22,327P/0F + reftest 712/712 zero inconsistent + clippy -D warnings + fmt clean. Residual beyond-minimum pool booked: replace-before-load deep gaps 35 / lazy dynamic-flip family 27 / the-iframe-element form faces 26 / nested navigation control 4T (skip pool) / popup management (user-activation infra). | html/browsers / history / navigation-api | 473 / 584 subtests | 81% | Completed | Oct 21, 2026 | GitHub ↗ |
| Web Crypto (subtle/random)Goal closed (DC-1~3 met) at 25788/33109 = 77.89% subtests; 73/123 case files all-green; non-tentative fails zero (remaining 7320 fails all tentative draft domains — ML-KEM/AES-OCB/Ed448/X448/KMAC/ML-DSA/ChaCha20-Poly1305/SHA3-XOF/Argon2 etc; reopen = upstream promotion per domain); single pre-existing iframe-dependent Timeout parked with DOM domain. M1->M6 six rounds 6728->25788 subtest-level zero regression; DC-3 gates: cargo fmt + guarded-clippy clean; make test all legs green; make reftest 712/712 (Layout 508/508 + Text 204/204). | WebCryptoAPI | 25,788 / 33,109 subtests | 77.9% | Completed | Oct 17, 2026 | GitHub ↗ |
| Web Crypto phase 2 (expansion + tentative promotion screening)Goal-26 phase-2 closeout: DC-1 corpus completion (202 upstream blobs, 198 files on-disk, metadata split-outs correct); DC-2 promotion screening (all phase-1 7320F parked domains promoted on upstream master cb6cc97a; D1 per-domain proposals 109 cases/18052 subtests handed to user, pending, non-blocking); DC-3 expansion baseline 25788P = 77.89% with status-account zero change across all absorb windows; DC-4 coverage re-test bit-identical to official snapshot (12541P delta + 265 uncovered files bucket-attributed). Zero-regression anchor re-verified x3 at closure tip f97190c41: RUN1/RUN3 byte-identical to phase-1 M1 anchor (7208436 B, sha 40ea3c8e), RUN2 W4 sample (sha 7cdea628) structured status account zero-diff 25788P+7320F+1T. Engine baseline closed via D2 trigger: PR #267 (D4 dual-surface adjudication li=list-item/canvas=inline) merged f97190c41, ua_display_li 1F cleared both surfaces; gate green except 4 attributed reds outside goal domain (compositor dmabuf 3F environmental /dev/dri-absent double-precedented; default_actions + mutation_arm deterministic-today pre-existing same-day accounts, merge diff zero touch-face). | WebCryptoAPI | 25,788 / 33,109 subtests | 77.9% | Completed | Nov 1, 2026 | GitHub ↗ |
| Compression StreamsGoal closed at 245/318 (77.04%); non-brotli surface 245/245 stable across M2/M3 (DC-2 zero decline two consecutive rounds); all 73 failures = mission-deferred brotli bucket (reopen condition: brotli crate dependency review). | compression | 245 / 318 subtests | 77% | Completed | Oct 15, 2026 | GitHub ↗ |
| Timing & Web AnimationsPost-close drift check on combined main (~350 engine commits from parallel flows after the M4 close): hr-time 11/16 + performance-timeline 36/44 + user-timing 561/570 + web-animations 502/959 (run-to-run variance +-2 (first run 503/958)). Zero net regression vs close 1107/1571: performance-timeline and user-timing identical subtest-for-subtest; hr-time +1P; the single web-animations flip (playback-rate progress) A/B-attributed to the navigation-compat M2-S5 pagereveal wrapper adding ~0.2ms between timeline reads (belongs to the bookkept instant-complete residual family); denominator growth is executable-face composition drift - not capability regression. | hr-time / performance-timeline / user-timing / web-animations | 1,110 / 1,589 subtests | 69.9% | Completed | Oct 11, 2026 | GitHub ↗ |
| Keyboard default actionsCompleted = goal's semantic scope closed; 7 failures + 2 timeouts counted in denominator with residual assertions attributed cross-stream. | keyboard | 18 / 27 tests | 66.7% | Completed | Sep 8, 2026 | GitHub ↗ |
| Performance API (resource/navigation/event/longtask)M3-S2 resource net metadata bridge: resource-timing 35/113 -> 42/113 (37.2%; __zwRtMeta carried on response object - encodedBodySize content-length-first non-consuming read / TextEncoder body bytes; decodedBodySize = encoded (identity local serving); transferSize = headers wire length + body bytes; nextHopProtocol http/1.1 (runner internal serving semantics); failed responses carry no meta -> size 0 per spec). make test 71 targets green. Cumulative pass 42+81+55+13=191. | resource-timing / navigation-timing / event-timing / longtask-timing | 191 / 293 subtests | 65.2% | Completed | Oct 13, 2026 | GitHub ↗ |
| Cookies (document.cookie)Goal close at M3 verdict (DC-1~3 all satisfied): executable face 533/533 = 100% (trajectory 3.9% -> 100%; monotonic no down-round); non-executable face 304 unchanged (303 zw-static-corpus marked + 1 transport-leg timeout). Gates: make reftest 708/708 zero regression; product-smoke welcome oracle diff 15.40% <= 20% byte-stable; make test decomposed legs all green incl. two cross-stream pre-existing reds root-caused and fixed during close (e776d6875 dropped render_session init -> cb4e15955, webdriver navigate family 8 red -> 14/14 green; 3ed4223cf duplicated #[test] slip -> 9ac80aedb, quickjs clippy leg compile red -> clean). One pre-existing foreign red on the record: zero-browser cdp_focus_governance quickjs leg (red at session start ddc0bf0f3, focus-governance surface, charged to focus stream per collision clause). Excluded faces remain charged: CookieStore API, SameSite/CHIPS/3PC, network transport (net jar domain). | cookies | 533 / 837 subtests | 63.7% | Completed | Oct 12, 2026 | GitHub ↗ |
| Intersection/Resize observersIntersectionObserver 100/218 (45.9%) + ResizeObserver 18/57 (31.6%) combined; residual clusters tracked as cross-stream coordination items. | intersection-observer / resize-observer | 118 / 275 subtests | 42.9% | Completed | Sep 12, 2026 | GitHub ↗ |
| Security (CSP / mixed content / secure contexts)74 of 445 CSP test files fully green at goal close (R11 verdict); CSP and Mixed Content enforcement default-on; mixed-content and secure-contexts gates verified via in-tree test gates. | content-security-policy / mixed-content / secure-contexts | 174 / 604 subtests | 28.8% | Completed | Sep 25, 2026 | GitHub ↗ |
| Coverage gap facets batch1 (twelve small suites)Twelve gap facets wired into resident channels (goal M1-M3). Denominator = local pinned-corpus (31597693) window-executable subtests only; content-skip faces (ref/manual/iframe/worker/non-testharness) and non-fetched wrapper faces (gen/ 1185+512; .any.js) excluded per M0 triage. Per-face P/total: quirks 35/75; domparsing 151/1217; domxpath 21/124; SRI 32/78; cors 92/346; import-maps 34/80; mixed-content 0/11; referrer-policy 1/41 (.py endpoint gap); input-events 188/1003 (editing-pipeline cluster triaged out); preload 106/407; wai-aria 2/424 (testdriver computed label/role gap); storage 0/5 (StorageManager engine gap). Zero-regression gates = per-face dry-run P (M1/M3 evidence). | quirks / domparsing / domxpath / subresource-integrity / cors / import-maps / mixed-content / referrer-policy / input-events / preload / wai-aria / storage | 662 / 3,811 subtests | 17.4% | Completed | Nov 6, 2026 | GitHub ↗ |
| CSS testharness all-directory expansion (phase 2)Phase-2 closeout: 44 batch directories wired into the resident channel (dry-run baselines B1 31.8% / B2 44.1% / B3 24.0%, 9,642P = 3.7x gate quota); anchor-size() argument-order canonical serialization fixed in place (+2,252, channel 40.6%->58.2%); phase-1 sixteen-dir zero-regression A/B (385 subtest regressions set-level attributed to pre-existing main evolution, booked to counterpart); DC-4 gate passed: css topdir 17,923 -> 29,514 = 15.09% (gate >=20,500 at 1.44x), full headline conversion 68.08% no regression; parked: layout goal proposal (>10k trigger confirmed, pending ruling), WAAPI/unimplemented API domains, cssom-view/geometry coordination face (M0 triage table). | css-values / selectors / css-variables / css-color / css-align / css-sizing / css-typed-om / css-fonts / css-shapes / css-anchor-position / css-masking / css-backgrounds / css-transforms / motion / css-grid / css-text / cssom / css-images / filter-effects / css-ui / css-borders / css-logical / css-text-decor / css-box / css-pseudo / css-display / css-mixins / css-syntax / fill-stroke / css-shadow / css-size-adjust / css-content / css-will-change / css-color-adjust / css-easing / compositing / css-counter-styles / css-nesting / css-color-hdr / css-env / css-conditional / mediaqueries / css-cascade / CSS2 / css-forms / css-link-params / css-font-loading / css-gaps / css-flexbox / css-multicol / css-position / css-overflow / css-tables / css-lists / css-inline / css-scroll-snap / css-viewport / css-break / css-contain / css-writing-modes | 29,514 / 195,553 subtests | 15.1% | Completed | Nov 5, 2026 | GitHub ↗ |
Weekly performance run
Performance at p95
These results come from the recorded GitHub runner class. Compare the same metric, unit and measurement conditions; a class label does not guarantee identical hardware. Lower is better.
Selected metric measurements
| Date | Result |
|---|---|
| Aug 8, 2026 | 229.88 ms |
| Aug 10, 2026 | 142.08 ms |
| Aug 13, 2026 | 318.97 ms |
| Aug 14, 2026 | 388.76 ms |
| Aug 16, 2026 | 1,036.45 ms |
| Aug 17, 2026 | 350.45 ms |
| Aug 18, 2026 | 215.81 ms |
| Aug 19, 2026 | 161.17 ms |
| Aug 20, 2026 | 218.79 ms |
| Aug 21, 2026 | 244.15 ms |
| Aug 22, 2026 | 180.96 ms |
| Aug 23, 2026 | 237.7 ms |
| Aug 24, 2026 | 223.59 ms |
| Aug 25, 2026 | 158.14 ms |
| Aug 26, 2026 | 230.05 ms |
| Aug 27, 2026 | 249.45 ms |
| Aug 28, 2026 | 222.65 ms |
| Aug 30, 2026 | 250.33 ms |
| Aug 31, 2026 | 269.7 ms |
| Sep 1, 2026 | 284.67 ms |
| Sep 2, 2026 | 244.78 ms |
| Sep 3, 2026 | 162.55 ms |
| Sep 4, 2026 | 219.76 ms |
| Sep 5, 2026 | 216.07 ms |
| Sep 6, 2026 | 278.32 ms |
| Sep 7, 2026 | 221.26 ms |
| Sep 8, 2026 | 276.68 ms |
| Sep 9, 2026 | 297.32 ms |
| Sep 12, 2026 | 280.45 ms |
| Sep 13, 2026 | 217.54 ms |
| Sep 14, 2026 | 268.2 ms |
| Sep 15, 2026 | 352.15 ms |
| Sep 16, 2026 | 231.98 ms |
| Sep 17, 2026 | 343.06 ms |
| Sep 18, 2026 | 216.74 ms |
| Sep 19, 2026 | 353.13 ms |
| Sep 27, 2026 | 373.1 ms |
| Sep 28, 2026 | 372.38 ms |
| Sep 29, 2026 | 374.41 ms |
| Sep 30, 2026 | 357.24 ms |
| Oct 2, 2026 | 381.42 ms |
| Oct 3, 2026 | 367.82 ms |
| Oct 4, 2026 | 385.61 ms |
| Oct 5, 2026 | 356.52 ms |
| Oct 6, 2026 | 384.61 ms |
| Oct 7, 2026 | 228.76 ms |
| Oct 8, 2026 | 368.98 ms |
| Oct 9, 2026 | 238.24 ms |
| Oct 10, 2026 | 410.24 ms |
| Oct 11, 2026 | 221.78 ms |
| Oct 12, 2026 | 401.6 ms |
| Oct 13, 2026 | 420.16 ms |
| Oct 14, 2026 | 352.98 ms |
| Oct 15, 2026 | 337.16 ms |
| Oct 16, 2026 | 237.84 ms |
| Oct 17, 2026 | 289 ms |
| Oct 18, 2026 | 433.54 ms |
| Oct 19, 2026 | 458.51 ms |
| Oct 23, 2026 | 346.53 ms |
| Oct 24, 2026 | 340.86 ms |
| Nov 1, 2026 | 513.03 ms |
| Nov 5, 2026 | 493.99 ms |
| Nov 6, 2026 | 509.59 ms |
| Nov 7, 2026 | 522.2 ms |
| Nov 11, 2026 | 522.43 ms |
Methodology: the CSS rendering suite is a self-source reftest — test and reference pages are both rendered by ZeroWeb and compared pixel by pixel, so it measures self-consistency, not a score judged by another browser. It must not be read as comparable to the full-suite scores Chrome, Firefox and Safari publish on wpt.fyi. The other suites are testharness runs scored by assertion.
Corpus version: suites run against a fixed imported snapshot of upstream WPT (CSS rendering currently v1.10). The snapshot is updated over time, so denominators differ between rows and between months; every count is shown as passed / total.
Reference points, for scale only (different methodology — not directly comparable): mainline browsers pass ~96–98% of the full WPT suite (wpt.fyi); the Ladybird independent engine reports ~93% (ladybird.org).
All benchmark histories
Welcome page (ms)
| Date | p95 |
|---|---|
| 2026-08-08 | 9.176914 ms |
| 2026-08-10 | 5.810163 ms |
| 2026-08-13 | 18.572647 ms |
| 2026-08-14 | 19.655495 ms |
| 2026-08-16 | 18.916102000000002 ms |
| 2026-08-17 | 12.731126 ms |
| 2026-08-18 | 10.673646999999999 ms |
| 2026-08-19 | 12.48348 ms |
| 2026-08-20 | 10.807693 ms |
| 2026-08-21 | 10.946962 ms |
| 2026-08-22 | 8.745001 ms |
| 2026-08-23 | 10.907375 ms |
| 2026-08-24 | 10.547122000000002 ms |
| 2026-08-25 | 7.6107 ms |
| 2026-08-26 | 10.8155 ms |
| 2026-08-27 | 11.229611 ms |
| 2026-08-28 | 10.493509 ms |
| 2026-08-30 | 11.224573 ms |
| 2026-08-31 | 11.630466 ms |
| 2026-09-01 | 11.617438 ms |
| 2026-09-02 | 11.27625 ms |
| 2026-09-03 | 8.805678 ms |
| 2026-09-04 | 9.343995000000001 ms |
| 2026-09-05 | 9.358026 ms |
| 2026-09-06 | 11.795615 ms |
| 2026-09-07 | 9.340326 ms |
| 2026-09-08 | 11.901821 ms |
| 2026-09-09 | 11.244878 ms |
| 2026-09-12 | 11.529263 ms |
| 2026-09-13 | 9.617934 ms |
| 2026-09-14 | 9.681208999999999 ms |
| 2026-09-15 | 12.86763 ms |
| 2026-09-16 | 10.381856 ms |
| 2026-09-17 | 13.123424 ms |
| 2026-09-18 | 7.981505 ms |
| 2026-09-19 | 13.082697000000001 ms |
| 2026-09-27 | 13.263309 ms |
| 2026-09-28 | 13.198992 ms |
| 2026-09-29 | 14.127039 ms |
| 2026-09-30 | 13.010849 ms |
| 2026-10-02 | 13.026494 ms |
| 2026-10-03 | 13.250195 ms |
| 2026-10-04 | 13.293546999999998 ms |
| 2026-10-05 | 13.052099 ms |
| 2026-10-06 | 13.640321 ms |
| 2026-10-07 | 7.811389 ms |
| 2026-10-08 | 13.131053 ms |
| 2026-10-09 | 8.306655 ms |
| 2026-10-10 | 13.446743000000001 ms |
| 2026-10-11 | 7.865056 ms |
| 2026-10-12 | 13.344033999999999 ms |
| 2026-10-13 | 13.716273 ms |
| 2026-10-14 | 11.00497 ms |
| 2026-10-15 | 12.397052 ms |
| 2026-10-16 | 9.066398 ms |
| 2026-10-17 | 9.41586 ms |
| 2026-10-18 | 15.819194000000001 ms |
| 2026-10-19 | 16.292586 ms |
| 2026-10-23 | 14.613111 ms |
| 2026-10-24 | 12.359555 ms |
| 2026-11-01 | 16.089689 ms |
| 2026-11-05 | 14.453657999999999 ms |
| 2026-11-06 | 14.395273999999999 ms |
| 2026-11-07 | 14.784479 ms |
| 2026-11-11 | 14.964395999999999 ms |
Medium page (ms)
| Date | p95 |
|---|---|
| 2026-08-08 | 229.875786 ms |
| 2026-08-10 | 142.075753 ms |
| 2026-08-13 | 318.966179 ms |
| 2026-08-14 | 388.760233 ms |
| 2026-08-16 | 1036.453199 ms |
| 2026-08-17 | 350.453732 ms |
| 2026-08-18 | 215.80676200000002 ms |
| 2026-08-19 | 161.173907 ms |
| 2026-08-20 | 218.794411 ms |
| 2026-08-21 | 244.15478 ms |
| 2026-08-22 | 180.96482 ms |
| 2026-08-23 | 237.698423 ms |
| 2026-08-24 | 223.58800300000001 ms |
| 2026-08-25 | 158.137103 ms |
| 2026-08-26 | 230.054858 ms |
| 2026-08-27 | 249.446719 ms |
| 2026-08-28 | 222.648615 ms |
| 2026-08-30 | 250.33177600000002 ms |
| 2026-08-31 | 269.704304 ms |
| 2026-09-01 | 284.67091 ms |
| 2026-09-02 | 244.78349599999999 ms |
| 2026-09-03 | 162.550157 ms |
| 2026-09-04 | 219.75535 ms |
| 2026-09-05 | 216.073325 ms |
| 2026-09-06 | 278.320218 ms |
| 2026-09-07 | 221.25514900000002 ms |
| 2026-09-08 | 276.67604 ms |
| 2026-09-09 | 297.319169 ms |
| 2026-09-12 | 280.45086299999997 ms |
| 2026-09-13 | 217.54214299999998 ms |
| 2026-09-14 | 268.20058 ms |
| 2026-09-15 | 352.146359 ms |
| 2026-09-16 | 231.983063 ms |
| 2026-09-17 | 343.05610199999995 ms |
| 2026-09-18 | 216.73845599999999 ms |
| 2026-09-19 | 353.133576 ms |
| 2026-09-27 | 373.097745 ms |
| 2026-09-28 | 372.375289 ms |
| 2026-09-29 | 374.40802099999996 ms |
| 2026-09-30 | 357.235099 ms |
| 2026-10-02 | 381.420643 ms |
| 2026-10-03 | 367.817279 ms |
| 2026-10-04 | 385.60855399999997 ms |
| 2026-10-05 | 356.517403 ms |
| 2026-10-06 | 384.61010899999997 ms |
| 2026-10-07 | 228.757954 ms |
| 2026-10-08 | 368.978306 ms |
| 2026-10-09 | 238.23639899999998 ms |
| 2026-10-10 | 410.23962800000004 ms |
| 2026-10-11 | 221.779215 ms |
| 2026-10-12 | 401.601672 ms |
| 2026-10-13 | 420.162866 ms |
| 2026-10-14 | 352.981279 ms |
| 2026-10-15 | 337.156097 ms |
| 2026-10-16 | 237.83626999999998 ms |
| 2026-10-17 | 289.000685 ms |
| 2026-10-18 | 433.540394 ms |
| 2026-10-19 | 458.51226 ms |
| 2026-10-23 | 346.532776 ms |
| 2026-10-24 | 340.86445100000003 ms |
| 2026-11-01 | 513.029904 ms |
| 2026-11-05 | 493.991804 ms |
| 2026-11-06 | 509.58904699999994 ms |
| 2026-11-07 | 522.197295 ms |
| 2026-11-11 | 522.429961 ms |
Morning test page (ms)
| Date | p95 |
|---|---|
| 2026-08-08 | 22.588928000000003 ms |
| 2026-08-10 | 17.45835 ms |
| 2026-08-13 | 85.124768 ms |
| 2026-08-14 | 98.125454 ms |
| 2026-08-16 | 78.85687800000001 ms |
| 2026-08-17 | 72.747853 ms |
| 2026-08-18 | 69.10854900000001 ms |
| 2026-08-19 | 53.300070999999996 ms |
| 2026-08-20 | 64.479394 ms |
| 2026-08-21 | 64.706025 ms |
| 2026-08-22 | 53.236479 ms |
| 2026-08-23 | 64.454052 ms |
| 2026-08-24 | 61.149243 ms |
| 2026-08-25 | 44.240954 ms |
| 2026-08-26 | 65.838572 ms |
| 2026-08-27 | 67.252315 ms |
| 2026-08-28 | 61.331826 ms |
| 2026-08-30 | 64.637701 ms |
| 2026-08-31 | 67.18861 ms |
| 2026-09-01 | 67.852128 ms |
| 2026-09-02 | 66.0274 ms |
| 2026-09-03 | 55.107728 ms |
| 2026-09-04 | 60.442744 ms |
| 2026-09-05 | 60.577684 ms |
| 2026-09-06 | 80.006734 ms |
| 2026-09-07 | 60.958774 ms |
| 2026-09-08 | 79.872874 ms |
| 2026-09-09 | 77.20222 ms |
| 2026-09-12 | 78.062799 ms |
| 2026-09-13 | 70.381299 ms |
| 2026-09-14 | 61.131118 ms |
| 2026-09-15 | 101.436745 ms |
| 2026-09-16 | 83.808092 ms |
| 2026-09-17 | 99.31234 ms |
| 2026-09-18 | 61.178151 ms |
| 2026-09-19 | 104.667712 ms |
| 2026-09-27 | 101.87996100000001 ms |
| 2026-09-28 | 105.47102899999999 ms |
| 2026-09-29 | 106.02719300000001 ms |
| 2026-09-30 | 105.025794 ms |
| 2026-10-02 | 102.08811499999999 ms |
| 2026-10-03 | 100.488998 ms |
| 2026-10-04 | 101.06187 ms |
| 2026-10-05 | 101.182978 ms |
| 2026-10-06 | 102.958582 ms |
| 2026-10-07 | 61.464328 ms |
| 2026-10-08 | 101.61581799999999 ms |
| 2026-10-09 | 62.64921 ms |
| 2026-10-10 | 102.199353 ms |
| 2026-10-11 | 61.190808 ms |
| 2026-10-12 | 117.83203999999999 ms |
| 2026-10-13 | 124.1177 ms |
| 2026-10-14 | 102.14187299999999 ms |
| 2026-10-15 | 95.594651 ms |
| 2026-10-16 | 72.37816699999999 ms |
| 2026-10-17 | 74.76153500000001 ms |
| 2026-10-18 | 126.24712499999998 ms |
| 2026-10-19 | 130.68232799999998 ms |
| 2026-10-23 | 119.29911899999999 ms |
| 2026-10-24 | 97.639047 ms |
| 2026-11-01 | 70.44588999999999 ms |
| 2026-11-05 | 70.793074 ms |
| 2026-11-06 | 70.77441800000001 ms |
| 2026-11-07 | 72.69730200000001 ms |
| 2026-11-11 | 81.757857 ms |
Browser startup (ms)
| Date | p95 |
|---|---|
| 2026-08-08 | 341.405658 ms |
| 2026-08-10 | 110.336726 ms |
| 2026-08-13 | 147.586293 ms |
| 2026-08-14 | 157.18157599999998 ms |
| 2026-08-16 | 92.198441 ms |
| 2026-08-17 | 50.550191 ms |
| 2026-08-18 | 46.768363 ms |
| 2026-08-19 | 126.653099 ms |
| 2026-08-20 | 41.894586999999994 ms |
| 2026-08-21 | 43.476749000000005 ms |
| 2026-08-22 | 108.600052 ms |
| 2026-08-23 | 100.80454599999999 ms |
| 2026-08-24 | 30.274803 ms |
| 2026-08-25 | 294.615162 ms |
| 2026-08-26 | 135.729885 ms |
| 2026-08-27 | 79.51859 ms |
| 2026-08-28 | 31.177815000000002 ms |
| 2026-08-30 | 32.753114 ms |
| 2026-08-31 | 33.990443 ms |
| 2026-09-01 | 41.851262 ms |
| 2026-09-02 | 136.912628 ms |
| 2026-09-03 | 230.06373299999998 ms |
| 2026-09-04 | 114.389912 ms |
| 2026-09-05 | 133.309525 ms |
| 2026-09-06 | 121.54630900000001 ms |
| 2026-09-07 | 30.629986000000002 ms |
| 2026-09-08 | 55.447738 ms |
| 2026-09-09 | 84.104113 ms |
| 2026-09-12 | 47.086220999999995 ms |
| 2026-09-13 | 27.875936 ms |
| 2026-09-14 | 303.721379 ms |
| 2026-09-15 | 390.412421 ms |
| 2026-09-16 | 31.488158000000002 ms |
| 2026-09-17 | 53.592974 ms |
| 2026-09-18 | 35.299261 ms |
| 2026-09-19 | 75.389185 ms |
| 2026-09-27 | 137.92918600000002 ms |
| 2026-09-28 | 119.01902100000001 ms |
| 2026-09-29 | 36.885296 ms |
| 2026-09-30 | 37.386056 ms |
| 2026-10-02 | 57.06465 ms |
| 2026-10-03 | 36.884018999999995 ms |
| 2026-10-04 | 38.587829 ms |
| 2026-10-05 | 37.762077999999995 ms |
| 2026-10-06 | 38.136849 ms |
| 2026-10-07 | 39.471419 ms |
| 2026-10-08 | 36.326657999999995 ms |
| 2026-10-09 | 25.783599000000002 ms |
| 2026-10-10 | 37.420249000000005 ms |
| 2026-10-11 | 97.152644 ms |
| 2026-10-12 | 37.510929 ms |
| 2026-10-13 | 169.63885100000002 ms |
| 2026-10-14 | 113.03706299999999 ms |
| 2026-10-15 | 74.150133 ms |
| 2026-10-16 | 40.006801 ms |
| 2026-10-17 | 32.070924 ms |
| 2026-10-18 | 40.188198 ms |
| 2026-10-19 | 226.464733 ms |
| 2026-10-23 | 46.100361 ms |
| 2026-10-24 | 39.990908999999995 ms |
| 2026-11-01 | 40.259429 ms |
| 2026-11-05 | 42.784907 ms |
| 2026-11-06 | 124.830056 ms |
| 2026-11-07 | 69.69612 ms |
| 2026-11-11 | 42.452833000000005 ms |
Peak memory (MB)
| Date | p95 |
|---|---|
| 2026-08-08 | 175.22265625 MB |
| 2026-08-10 | 180.5078125 MB |
| 2026-08-13 | 187.37890625 MB |
| 2026-08-14 | 152.74609375 MB |
| 2026-08-16 | 167.61328125 MB |
| 2026-08-17 | 131.43359375 MB |
| 2026-08-18 | 130.703125 MB |
| 2026-08-19 | 128.46484375 MB |
| 2026-08-20 | 130.48046875 MB |
| 2026-08-21 | 130.40234375 MB |
| 2026-08-22 | 130.9296875 MB |
| 2026-08-23 | 130.74609375 MB |
| 2026-08-24 | 131.33203125 MB |
| 2026-08-25 | 130.74609375 MB |
| 2026-08-26 | 131.046875 MB |
| 2026-08-27 | 131.36328125 MB |
| 2026-08-28 | 132.02734375 MB |
| 2026-08-30 | 134.9921875 MB |
| 2026-08-31 | 135.32421875 MB |
| 2026-09-01 | 135.421875 MB |
| 2026-09-02 | 135.54296875 MB |
| 2026-09-03 | 138.48046875 MB |
| 2026-09-04 | 137.99609375 MB |
| 2026-09-05 | 138.12890625 MB |
| 2026-09-06 | 138.0703125 MB |
| 2026-09-07 | 138.47265625 MB |
| 2026-09-08 | 137.80859375 MB |
| 2026-09-09 | 138.921875 MB |
| 2026-09-12 | 142.55078125 MB |
| 2026-09-13 | 145.16796875 MB |
| 2026-09-14 | 145.9921875 MB |
| 2026-09-15 | 148.33203125 MB |
| 2026-09-16 | 148.0078125 MB |
| 2026-09-17 | 148.09375 MB |
| 2026-09-18 | 148.2265625 MB |
| 2026-09-19 | 149.75 MB |
| 2026-09-27 | 152.41015625 MB |
| 2026-09-28 | 151.921875 MB |
| 2026-09-29 | 152.671875 MB |
| 2026-09-30 | 152.6484375 MB |
| 2026-10-02 | 152.62109375 MB |
| 2026-10-03 | 152.71484375 MB |
| 2026-10-04 | 152.8671875 MB |
| 2026-10-05 | 153.10546875 MB |
| 2026-10-06 | 152.62890625 MB |
| 2026-10-07 | 153.0 MB |
| 2026-10-08 | 152.734375 MB |
| 2026-10-09 | 153.01953125 MB |
| 2026-10-10 | 153.359375 MB |
| 2026-10-11 | 153.625 MB |
| 2026-10-12 | 153.59765625 MB |
| 2026-10-13 | 153.4296875 MB |
| 2026-10-14 | 153.92578125 MB |
| 2026-10-15 | 154.40234375 MB |
| 2026-10-16 | 154.15234375 MB |
| 2026-10-17 | 154.4140625 MB |
| 2026-10-18 | 155.78515625 MB |
| 2026-10-19 | 157.2890625 MB |
| 2026-10-23 | 157.80078125 MB |
| 2026-10-24 | 159.55859375 MB |
| 2026-11-01 | 143.796875 MB |
| 2026-11-05 | 143.8984375 MB |
| 2026-11-06 | 145.13671875 MB |
| 2026-11-07 | 145.29296875 MB |
| 2026-11-11 | 147.515625 MB |
Engineering guides
How ZeroWeb tests AI-autonomous software engineering
A browser gives AI development a long-running test case: architecture, code, debugging and verification must work together.
ZeroWeb Rust browser architecture: from request to pixels
Follow the page pipeline and the boundaries between the browser application, embeddable WebView, engine and rendering infrastructure.
Getting started with ZeroWebView and ZeroBrowser
Prepare a local checkout, launch the desktop browser and inspect the WebView API before embedding the experimental engine.
How to read ZeroWeb WPT and performance results
Understand self-source reftests, assertion-based suites, changing corpus sizes and p95 benchmarks before comparing results.