google/gemma-2-9b-it

vllm 0.21.0 on 8x NVIDIA H100 80GB HBM3 (fp16) · seed 42 · run 019e3b97-3b4f-73a5-84d6-49db6f8586b4 · signed

Headline metrics

TTFT P50 (ms)	30.05
TTFT P99 (ms)	184
Throughput (tok/s)	385
$/M tokens	—
J/token	2.31
Power avg (W)	901
Power peak (W)	938
WER mean	—
J / audio s	—

All metrics

compliance_rate	0.9861
energy_joules_total	18,851
joules_per_token	2.31
ok_rate	1.00
power_avg_w	901
power_peak_w	938
req_per_s_all	3.39
req_per_s_passing	3.35
slo_hardware_class	h100
slo_template_resolved	ttft<200ms, tpot<50ms, total<3000ms
throughput_tok_per_s	385
total_p50_ms	1,095
total_p99_ms	1,423
tpot_p50_ms	8.65
tpot_p99_ms	10.80
ttft_p50_ms	30.05
ttft_p99_ms	184

Provenance

Model revision	`unknown00`
Provider	vllm
Hardware fingerprint	`550474fc9132129654f5d20c316eaffec99a1f67c0cc0b4641e60efb46ea79e7`
Dataset	builtin-chatbot-short · `6f4b1f68fc3a813baa983cbe70cd9ef57f8c86e6b2e6ccc9aaa2a498e588d510`
Timestamp	2026-05-18T14:57:07.407582+00:00
Envelope JSON	f0b1a3bfe520.json

Verify this result

bench verify /bench/leaderboard/envelopes/f0b1a3bfe520.json

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