InferenceBench

Qwen/Qwen2.5-7B-Instruct

vllm unknown on 8x NVIDIA H100 80GB HBM3 (bf16) · seed 0 · run 019ebfa5-3db5-787a-b521-841576ef76e7 · signed

Headline metrics

TTFT P50 (ms) 13.47
TTFT P99 (ms)
Throughput (tok/s)
$/M tokens
J/token 6.33
Power avg (W) 881
Power peak (W) 935
WER mean
J / audio s

All metrics

energy_joules_total 8,113
joules_per_token 6.33
n_ok 5.00
n_samples 5.00
ok_rate 1.00
pass_at_1 1.00
pass_at_1_p05 1.00
pass_at_1_p50 1.00
pass_at_1_p95 1.00
power_avg_w 881
power_peak_w 935
timeout_rate 0.0000
tokens_out_total 1,281
total_p50_ms 1,699
ttft_p50_ms 13.47

Provenance

Model revisionunknown00
Providervllm
Hardware fingerprintc353425077b20b36c71d9c32d9e5d49096478f4a0f7d6078c1dfc83c408ccf50
Datasetbuiltin-humaneval-mini · c9122e6eca7b960ed8a8dc8fcb40702b25024622a3ab90c546914b949193da06
Timestamp2026-06-13T06:22:18.037262+00:00
Envelope JSON9bde7a8bfedf.json

Verify this result

bench verify /bench/leaderboard/envelopes/9bde7a8bfedf.json

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