InferenceBench

Qwen/Qwen2.5-VL-7B-Instruct

vllm unknown on 8x NVIDIA H100 80GB HBM3 (bf16) · seed 0 · run 019ebfa8-ea53-7507-a36f-83efab831c3f · signed

Headline metrics

TTFT P50 (ms) 50.78
TTFT P99 (ms)
Throughput (tok/s)
$/M tokens
J/token 105
Power avg (W) 620
Power peak (W) 634
WER mean
J / audio s

All metrics

accuracy 0.8000
accuracy_p05 0.2000
accuracy_p50 1.00
accuracy_p95 1.00
energy_joules_total 1,363
joules_per_token 105
n_ok 5.00
n_samples 5.00
ok_rate 1.00
power_avg_w 620
power_peak_w 634
tokens_out_total 13.00
total_p50_ms 50.78
ttft_p50_ms 50.78

Provenance

Model revisionunknown00
Providervllm
Hardware fingerprintc353425077b20b36c71d9c32d9e5d49096478f4a0f7d6078c1dfc83c408ccf50
Datasetbuiltin-chart-qa-mini · 2057d9889a63d42a0fa66599cb0ca9b9233cb4f656dffe886705b4a03229a1b9
Timestamp2026-06-13T06:26:18.835234+00:00
Envelope JSON9bffffe6399d.json

Verify this result

bench verify /bench/leaderboard/envelopes/9bffffe6399d.json

bench verify re-downloads this envelope, recomputes the canonical content hash, and validates the Sigstore signature against the embedded certificate + Rekor entry.