Qwen/Qwen2.5-7B-Instruct

vllm 0.21.0 on 8x NVIDIA H100 80GB HBM3 (fp16) · seed 42 · run 019e3b39-5d04-7afb-a506-fe3140835575 · signed

Headline metrics

TTFT P50 (ms)	26.84
TTFT P99 (ms)	52.98
Throughput (tok/s)	541
$/M tokens	—
J/token	1.65
Power avg (W)	909
Power peak (W)	948
WER mean	—
J / audio s	—

All metrics

compliance_rate	0.9896
energy_joules_total	18,543
joules_per_token	1.65
ok_rate	1.00
power_avg_w	909
power_peak_w	948
req_per_s_all	4.63
req_per_s_passing	4.58
slo_hardware_class	h100
slo_template_resolved	ttft<200ms, tpot<50ms, total<3000ms
throughput_tok_per_s	541
total_p50_ms	822
total_p99_ms	929
tpot_p50_ms	6.27
tpot_p99_ms	6.50
ttft_p50_ms	26.84
ttft_p99_ms	52.98

Provenance

Model revision	`unknown00`
Provider	vllm
Hardware fingerprint	`550474fc9132129654f5d20c316eaffec99a1f67c0cc0b4641e60efb46ea79e7`
Dataset	builtin-chatbot-short · `6f4b1f68fc3a813baa983cbe70cd9ef57f8c86e6b2e6ccc9aaa2a498e588d510`
Timestamp	2026-05-18T13:14:35.652372+00:00
Envelope JSON	43a8c771ef48.json

Verify this result

bench verify /bench/leaderboard/envelopes/43a8c771ef48.json

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