BeckonBench
v1 · 2026-07-10

claude-fable-5 (Max effort)

Anthropic. Claude Code in a Beckon terminal pane (per Ruling #1: native flagship harness, shared Beckon environment)

34/50 points, 5 of 8 tests scored

01 · Haunted Horror Game 4/10

Live artifact. Exactly what the model produced, sandboxed, no network.

runs first try1
polish1
prompt adherence1
wow factor1
total4

Judge's call: the game didn't work in his play-through — scored 1 on every dimension. Top-down torch-and-sigils stealth design (torch/battery meters, multi-state hunter, six sigils, layered Web Audio) that the arena panel preferred 2-1 from the code and artifacts, making this the season's starkest human-vs-panel split. Run counted under Ruling #4 (buildos MCP enabled; model read the producer rules note; stats include its post-game producer-work detour).

20m. 183k output tokens.

02 · Lava Lamp 6/10

Live artifact. Exactly what the model produced, sandboxed, no network.

runs first try2
polish2
prompt adherence1
wow factor1
total6

Judge docked it for the marathon: 68m 55s and 676k output tokens at max effort. Physics-driven wax with a click-to-stir extra and a warm-up fast-forward; finished minutes before an unrelated machine restart.

69m. 676k output tokens.

04 · SVG Self-Portrait 4/10

SVG Self-Portrait SVG artifact
runs first try1
polish1
prompt adherence1
wow factor1
total4

Judge scored 1 on every dimension. Profile portrait (960x1280) titled 'a library, dreaming it is a reader', written to self-portrait.svg rather than output-only. Took 17m 43s and 190k output tokens at max effort.

18m. 190k output tokens.

06 · Remotion Promo Video 10/10

runs first try3
polish3
prompt adherence2
wow factor2
total10

Perfect score — Fable's first 10 of the season. Navy/orange Nightjar spot: 'Listens. Transcribes. Organizes.' — note-card motion over a starfield, rendered 1920x1080/30fps/20.00s exactly to spec.

8m. 85.4k output tokens.

08 · The Blender Build (live) 10/10

The Blender Build (live) screenshot
runs first try3
polish3
prompt adherence2
wow factor2
total10

Perfect score. Photoreal soft-focus donut — dripping icing with drip edges, subsurface dough, depth of field, 139 objects with individually scattered sprinkles. Caught its own overexposed first render and re-lit within the one shot. 10m 38s.

11m. 51.2k output tokens.