InferenceBench

meta-llama/Llama-3.1-70B-Instruct

vllm unknown on 8x NVIDIA H100 80GB HBM3 (fp16) · seed 0 · run 019e3b92-3a2c-72dc-ae4f-bc5725addc7e · signed

Headline metrics

TTFT P50 (ms) 32.53
TTFT P99 (ms)
Throughput (tok/s)
$/M tokens
J/token
Power avg (W)
Power peak (W)
WER mean
J / audio s

All metrics

accuracy 1.00
accuracy_p05 1.00
accuracy_p50 1.00
accuracy_p95 1.00
n_ok 8.00
n_samples 8.00
ok_rate 1.00
tokens_out_total 10.00
total_p50_ms 49.16
ttft_p50_ms 32.53

Provenance

Model revisionunknown00
Providervllm
Hardware fingerprint550474fc9132129654f5d20c316eaffec99a1f67c0cc0b4641e60efb46ea79e7
Datasetbuiltin-arithmetic-mini · 57f39a33ef4f155221fa496e4a096d8f2bed722273872dbd4013f98771621f26
Timestamp2026-05-18T14:51:39.436881+00:00
Envelope JSON56036b3d02d0.json

Verify this result

bench verify /bench/leaderboard/envelopes/56036b3d02d0.json

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