Cloud baseline

Hy4 flash — 0.5874 on Gauntlet 7.2

tencent/hy4-flash · scorer sb-7.2

Hy4 flash, single model via OpenRoutertencent/hy4-flash · single agentNov 17, 2026
Overall
0.5874
Excellent
no
Benchmark
Gauntlet 7.2
Model build
30m 54s
Prompt tok
14.3M
Gen tok
126.9k

Recorded inference cost

$0.05

From the saved OpenRouter generation billing summary (137 recorded requests). Whether auxiliary model calls were included was not retained. Scoring reuses the saved build; this amount describes the original inference spend.

Token usage

Tokens: 14.3M in · 127k out · 13.9M cached (exact, OpenRouter records)

14,451,453 tokens over 137 model calls: every call of the run, the model's own and the harness's helper calls, counted from OpenRouter records. Reasoning tokens as the provider reports them: 62,352 (not added to the total).

  • tencent/hy4-flash

    In 14.3M · Out 127k · Cached 13.9M · Reasoning 62.4k · 136 calls

  • google/gemini-2.5-flash

    In 974 · Out 5 · Cached 0 · Reasoning 0 · 1 call

Tier breakdown

A · structure & runtime1.00
B · behaviour1.00
C · vendor contract1.00
D · finesse0.97

Scoring detail

How this score was built

Per-check scorer output, as posted by the app. Expand a section to inspect its checks.

Earned credit before admission

(0.88 × 0.7910 core + 0.12 × 0.77 gate × 0.80 excellence) × 1.0000 critical = 0.7699

The core (88% of the total) is the weighted mean of the thirteen measured tiers. The last 12% is the excellence slice: it unlocks in proportion to the perfection conditions below (11 of 16 met here), then pays out at the excellence tier's own measured mean. Core 0.6961 + excellence 0.0739.

j_first_use 1.00j_workflow_journey 1.00j_error_state 0.30j_empty_state 1.00console_clean 1.00v_responsive_375 1.00v_dates_readable 1.00t_scene_binding 0.21x_conservation_residual 1.00r_no_row_loss 1.00p_drag_frames 1.00p_idle_flatness 0.80p_stream_apply 0.00p_under_stream 1.00p_api_latency 1.00p_sync_wall 1.00

Earned score 0.770 · Admission ceiling 0.599 · Final score 0.587

Final score is the lower of earned credit and the admission ceiling. Passing admission adds no points.

Visible: not metMatching: not metGood: not metVisual excellence: not metBackend excellence: passed

Visible data-backed 3D: s_visible_surface (maximum 0.599)

Tower structure, currency mapping and scene truth: s_visible_surface, s_tower_geometry, s_currency_collar, t_scene_binding, t_vs7dbg_truth (maximum 0.609)

3D interaction and overview legibility: t_coast_identity, t_coast_reality, t_labels_culling, t_brush_link, t_stream_diff, q_overview_legibility, q_inspector_framing (maximum 0.699)

Event animation and backend recovery: m_committed_event_replay (maximum 0.899)

Token rates

Measured by the engine itself, one record per completed model call: prefill rate is prompt tokens over time-to-first-token, generation rate is completion tokens over the decode window. Medians per node.

NodeCallsPrompt tokGen tokPrefill tok/sGen tok/s
gemini197452122.0166.7
hy413614,323,580126,89425015.9147.7
fleet13714,324,554126,89924935.8147.9

Graded browser recording

Loading duration…

Watch the full graded browser recording at its original speed: the payment field, structure inspection, committed updates and replay. The check results and screenshots on this page provide the wider evidence. Pause or scrub to inspect a view.

Each tower represents a payment: height encodes its amount, cap color its status, and the collar its currency. The selected detailed tower is expected to animate its collar after a committed backend update or replay. These are the task requirements; the check results show what this app actually achieved.

Ready to play0:00 / —:—
Open recording file

Drag the timeline to seek. Keyboard: arrow keys move 5 seconds; Home and End jump to the start and end.

Graded browser recording re-encoded to 760x476 at 25 fps to fit 4 MiB, same 201.4 s; graded original sha256 943ed0c75b5ab5544475c01dded9e081a4aaef880379ff04e42275ffc742f8a1

Screenshots

Captured from the built application during browser grading. Captions identify the recorded views and checks; these images do not represent repair rounds.

1First captured render
2Latest captured render
3After sync
4Error state
5Mobile · 375px

Run details

Model
tencent/hy4-flash
Engine events
0
Repair rounds
0
Started
Nov 17, 2026, 04:39 PM
Finished
Nov 17, 2026, 05:19 PM
Call budget
With wrap-up nudge, version 2: as the call budget ran low, the agent harness told the model to start no new work and finish

How the benchmark is scored