Home /
llm.inference.chatbot-short /
019ebf74-9c6e-7b36-9234-39c3fd8ced02
Qwen/Qwen2.5-72B-Instruct
vllm 0.22.1 on 8x NVIDIA H100 80GB HBM3
(bf16) · seed 42 · run 019ebf74-9c6e-7b36-9234-39c3fd8ced02
· signed
Headline metrics
TTFT P50 (ms)
24.21
TTFT P99 (ms)
33.59
Throughput (tok/s)
56.47
$/M tokens
—
J/token
37.18
Power avg (W)
2,112
Power peak (W)
2,189
WER mean
—
J / audio s
—
All metrics
compliance_rate
1.00
energy_joules_total
189,335
joules_per_token
37.18
ok_rate
1.00
power_avg_w
2,112
power_peak_w
2,189
req_per_s_all
0.4768
req_per_s_passing
0.4768
slo_hardware_class
h100
slo_template_resolved
ttft<200ms, tpot<50ms, total<3000ms
throughput_tok_per_s
56.47
total_p50_ms
2,062
total_p99_ms
2,073
tpot_p50_ms
16.05
tpot_p99_ms
16.39
ttft_p50_ms
24.21
ttft_p99_ms
33.59
Provenance
Model revision unknown00
Provider vllm
Hardware fingerprint c353425077b20b36c71d9c32d9e5d49096478f4a0f7d6078c1dfc83c408ccf50
Dataset builtin-chatbot-short · 6f4b1f68fc3a813baa983cbe70cd9ef57f8c86e6b2e6ccc9aaa2a498e588d510
Timestamp 2026-06-13T05:29:11.022500+00:00
Envelope JSON c1-79d354b13545.json
Verify this result
bench verify /bench/leaderboard/envelopes/c1-79d354b13545.json
bench verify re-downloads this envelope, recomputes the canonical
content hash, and validates the Sigstore signature against the embedded
certificate + Rekor entry.