Home /
llm.inference.chatbot-short /
019ebf76-0816-7135-b6a6-6b21ed5abb8d
Qwen/Qwen2.5-72B-Instruct
vllm 0.22.1 on 8x NVIDIA H100 80GB HBM3
(bf16) · seed 42 · run 019ebf76-0816-7135-b6a6-6b21ed5abb8d
· signed
Headline metrics
TTFT P50 (ms)
46.34
TTFT P99 (ms)
48.40
Throughput (tok/s)
234
$/M tokens
—
J/token
9.32
Power avg (W)
2,184
Power peak (W)
2,198
WER mean
—
J / audio s
—
All metrics
compliance_rate
1.00
energy_joules_total
200,158
joules_per_token
9.32
ok_rate
1.00
power_avg_w
2,184
power_peak_w
2,198
req_per_s_all
1.98
req_per_s_passing
1.98
slo_hardware_class
h100
slo_template_resolved
ttft<200ms, tpot<50ms, total<3000ms
throughput_tok_per_s
234
total_p50_ms
2,130
total_p99_ms
2,135
tpot_p50_ms
16.43
tpot_p99_ms
16.80
ttft_p50_ms
46.34
ttft_p99_ms
48.40
Provenance
Model revision unknown00
Provider vllm
Hardware fingerprint c353425077b20b36c71d9c32d9e5d49096478f4a0f7d6078c1dfc83c408ccf50
Dataset builtin-chatbot-short · 6f4b1f68fc3a813baa983cbe70cd9ef57f8c86e6b2e6ccc9aaa2a498e588d510
Timestamp 2026-06-13T05:30:44.118360+00:00
Envelope JSON c4-95cbfd4f49dc.json
Verify this result
bench verify /bench/leaderboard/envelopes/c4-95cbfd4f49dc.json
bench verify re-downloads this envelope, recomputes the canonical
content hash, and validates the Sigstore signature against the embedded
certificate + Rekor entry.