Measurements: the reference tests and the scripted subtests as they stand
Shadows; the gaps found on the way; the scores
Animations and transitions; the scores
Shadow trees, modules and import maps, classes that extend the platform's interfaces; the scores
Fonts fetched when text needs them; what a realm makes when; the instruments; the scores of today
Measurements: the harness and test262 figures of 2026-10-01
Measurements: the interface tests run and are counted
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The shell: a thread for every tab; the driver and the held-time line; the harness figure
Measurements: the harness figure from the rebuilt engine
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Measurements: the media directories join the harness; test262 41,607
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
The scripted harness figure is 135,632 of 159,192 subtests
Measurements: the scores of f1ca52e, and intl402, canvas, IndexedDB and css-display in the suites
Measurements: the scripted suite over 28 directories, and the scores of f1629b2
The scripted suite, measured over the wider list
The headline figures catch up with the engine: the CSS reference tests at
9458 of 14,616, and the scripted tests at 129,609 of 144,207 over the
twenty-six directories the suite now covers rather than the seventeen it
named.
That widening is worth its own paragraph here, because the number moved for
two different reasons and a reader deserves to know which. Nine directories
were added deliberately: a score measured only where the engine had already
built something measures the engine's own work, not the web, and the honest
figure over the wider list began at 78.7 per cent.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Fixed boxes: placed in turn, kept under their stacking context, clipped by the viewport alone
Pictures: height:auto over the attribute, aspect-ratio, object-fit; one pointer move a turn; the reftest score
Pointing in the painter's order, the viewport's reach, the document base URL, pushState reaching the shell, the traversal interfaces; the scripted score
Frames the reader can use, framesets, window.open, and the scripted score at 65822 of 74294
Scripting and Measurements: ranges, CDATA sections and processing instructions; 65627 / 74294 scripted subtests
Measurements: 17352 / 26275 scripted subtests, 9427 / 14616 reference tests
An object's and an embed's documents, pictures and fallback; 16886 / 26275 scripted, 9410 / 14616 reference
A frame element's window, names fixed at creation and window reuse; 16828 / 26275 scripted
The scripted tests: 16773 / 26154
A frame's window through its WindowProxy, navigation by its Location, the sandbox attribute and structured clone; 16739 / 26149 scripted, 9409 / 14616 reference
A frame's document runs its scripts and follows its iframe; the scripted WPT suite scored, 15326 / 26067
Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ebXZs43Ks9pd5MerR62rG
Frames drawn from their live documents, with the hosts answering for them; WPT 9405
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ebXZs43Ks9pd5MerR62rG
A float's percentage height resolves, and out-of-flow margins stay out; WPT 9403
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ebXZs43Ks9pd5MerR62rG
Array.from and Array.of construct their receiver; more operations keep to their realm; test262 40358
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_011ebXZs43Ks9pd5MerR62rG