Blackhole p150 tt-metal: Whisper distil-large-v3 batch-1 tok/s per user
263.4 tok/s per user at batch 1 (equals aggregate at batch 1); vendor marks Blackhole numbers as still-optimizing
Fields
id
benchmarks/ttmetal-p150-whisper-distil-large-v3-b1-tpsuMetric
tok_s_per_user
Value
263.4
Unit
tok/s per user at batch 1 (equals aggregate at batch 1); vendor marks Blackhole numbers as still-optimizing
Model
Whisper distil-large-v3
Methodology
VENDOR demo in tenstorrent/tt-metal models/README.md speech-to-text table (matrix last updated 2025-12-08). Whisper distil-large-v3, single p150 (Blackhole), batch 1, TT-Metalium v0.65.0-dev20251208. t/s/u 263.4 = 1/inter-token-latency; at batch 1 aggregate t/s is also 263.4. Blackhole rows in this matrix carry the vendor's caveat that the software is under active development, so treat as a moving target.
Measured by
vendor
Reproducible
yes
notes
Vendor demo numbers, not a vLLM serving trace. The README states t/s = t/s/u * batch and that vLLM-server performance may differ. LLM rows use input sequence length 128 except Mamba, which accepts any length. Starred Blackhole figures carry the vendor's own caveat that the software is still under active optimization.
Notes
Vendor demo numbers, not a vLLM serving trace. The README states t/s = t/s/u * batch and that vLLM-server performance may differ. LLM rows use input sequence length 128 except Mamba, which accepts any length. Starred Blackhole figures carry the vendor's own caveat that the software is still under active optimization.