Qwen/Qwen2-VL-7B-Instruct

vllm 0.21.0 on 8x NVIDIA H100 80GB HBM3 (fp16) · seed 42 · run 019e3b85-e34e-706c-826e-d44636d028c2 · signed

Headline metrics

TTFT P50 (ms)	42.50
TTFT P99 (ms)	198
Throughput (tok/s)	216
$/M tokens	—
J/token	3.83
Power avg (W)	837
Power peak (W)	857
WER mean	—
J / audio s	—

All metrics

compliance_rate	0.9767
energy_joules_total	18,005
joules_per_token	3.83
ok_rate	1.00
power_avg_w	837
power_peak_w	857
req_per_s_all	1.97
req_per_s_passing	1.93
slo_hardware_class	h100
slo_template_resolved	ttft<200ms, tpot<50ms, total<3000ms
throughput_tok_per_s	216
total_p50_ms	1,981
total_p99_ms	2,016
tpot_p50_ms	15.35
tpot_p99_ms	15.80
ttft_p50_ms	42.50
ttft_p99_ms	198

Provenance

Model revision	`unknown00`
Provider	vllm
Hardware fingerprint	`550474fc9132129654f5d20c316eaffec99a1f67c0cc0b4641e60efb46ea79e7`
Dataset	builtin-chatbot-short · `6f4b1f68fc3a813baa983cbe70cd9ef57f8c86e6b2e6ccc9aaa2a498e588d510`
Timestamp	2026-05-18T14:38:10.766378+00:00
Envelope JSON	835fae0e126c.json

Verify this result

bench verify /bench/leaderboard/envelopes/835fae0e126c.json

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