Task speech-to-text · model-size class Whisper base · batch 1 · 5 platforms · latency ≤ 50 ms treated as ≥ 20.00 FPS (single-stream approximation)
✓ Like-for-like: same model-size class and batch.
| Platform | Result | Model · precision · stack | Grade | Power | Price | Fit perf/power/budget | Source |
|---|---|---|---|---|---|---|---|
| Qualcomm Snapdragon X2 Elite (CRD) | 408 FPS inference_time b=1 | whisper_base / decoder float · Qualcomm AI Hub precompiled_qnn_onnx (QAIRT 2.50.0.260828221209) | D | unknown no power published | unknown | yes / – / – | Qualcomm AI Hub Models perf.yaml (BSD-3) |
| Qualcomm Snapdragon X Elite (CRD) | 272 FPS inference_time b=1 | whisper_base / decoder float · Qualcomm AI Hub precompiled_qnn_onnx (QAIRT 2.50.0.260828221209) | D | unknown no power published | unknown | yes / – / – | Qualcomm AI Hub Models perf.yaml (BSD-3) |
| Hailo-8L (M.2/PCIe, x86 host) | 41.80 FPS throughput b=1 | whisper_base_5s_no_kqs_decoder INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) | D | 1.50 W reported in result source — spec | unknown | yes / – / – | Hailo Model Zoo public models (Hailo-8L) |
| Qualcomm Dragonwing IQ-9075 EVK | 196 FPS inference_time b=1 | whisper_base / decoder float · Qualcomm AI Hub qnn_context_binary (QAIRT 2.50.0.260828221209) | D | unknown no power published | unknown | yes / – / – | Qualcomm AI Hub Models perf.yaml (BSD-3) |
| Hailo-8 (M.2/PCIe, x86 host) | 51.40 FPS throughput b=1 | whisper_base_5s_encoder INT8 (Hailo quantized) · Hailo DFC v2.20.0 / HailoRT (model zoo) | D | 2.50 W reported in result source — spec | unknown | yes / – / – | Hailo Model Zoo public models (Hailo-8) |
Confidence: A audited MLPerf closed, B documented third-party, C community, D vendor-published, E estimate. Unknown is never assumed. Other like-for-like groups: Whisper base b1 (5); Whisper tiny b1 (5); Whisper small b1 (4); Whisper large b1 (3); Whisper base b8 (2); Whisper tiny b8 (2); RNN-T bunknown (3); unknown bunknown (48)