Systran/faster-whisper-large-v3

whisper-http unknown on 8x NVIDIA H100 80GB HBM3 (fp16) · seed 0 · run 019e7488-e07c-74cd-b65e-52154b0c2eb6 · signed

Headline metrics

TTFT P50 (ms)	—
TTFT P99 (ms)	—
Throughput (tok/s)	—
$/M tokens	—
J/token	—
Power avg (W)	584
Power peak (W)	681
WER mean	0.0700
J / audio s	1,357

All metrics

audio_path_resolved_count	5.00
energy_joules_total	26,768
joules_per_audio_second	1,357
n_ok	5.00
n_samples	5.00
ok_rate	1.00
power_avg_w	584
power_peak_w	681
total_audio_duration_s	19.73
total_p50_ms	262
wer_mean	0.0700
wer_p50	0.0000
wer_p95	0.2200

Provenance

Model revision	`unknown00`
Provider	whisper-http
Hardware fingerprint	`c353425077b20b36c71d9c32d9e5d49096478f4a0f7d6078c1dfc83c408ccf50`
Dataset	librispeech-clean-mini · `cff1fdb61fe89ecdd453b7bbcc7850160645af4bc819dde51cd8dfcdf9c4eb33`
Timestamp	2026-05-29T16:19:47.964753+00:00
Envelope JSON	f1c81b4978c8.json

Verify this result

bench verify /bench/leaderboard/envelopes/f1c81b4978c8.json

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