InferenceBench

Qwen/Qwen2.5-7B-Instruct

vllm unknown on 8x NVIDIA H100 80GB HBM3 (fp16) · seed 0 · run 019e3b39-b322-7a92-aad2-b6166fcb8126 · signed

Headline metrics

TTFT P50 (ms) 14.26
TTFT P99 (ms)
Throughput (tok/s)
$/M tokens
J/token
Power avg (W)
Power peak (W)
WER mean
J / audio s

All metrics

accuracy 0.9200
accuracy_p05 0.6800
accuracy_p50 1.00
accuracy_p95 1.00
drift_rate 0.2000
mean_drift_turn 2.00
n_ok 5.00
n_samples 5.00
ok_rate 1.00
persona_consistency_mean 0.9200
persona_consistency_p50 1.00
persona_consistency_p95 1.00
tokens_out_total 1,124
total_p50_ms 1,732
ttft_p50_ms 14.26

Provenance

Model revisionunknown00
Providervllm
Hardware fingerprint550474fc9132129654f5d20c316eaffec99a1f67c0cc0b4641e60efb46ea79e7
Datasetpersona-consistency-mini · ac864a018c1eb985c7f697a997162da20039e99fff11927ec3139fa4f2d4e9e3
Timestamp2026-05-18T13:14:57.698521+00:00
Envelope JSON6d91ef5acf07.json

Verify this result

bench verify /bench/leaderboard/envelopes/6d91ef5acf07.json

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