Automatic Speech Recognition on AISHELL Eval A 2021 (CER)
3.12CERNIM4-ASR
Evaluation Results
| Method | Links | |
|---|---|---|
| NIM4-ASRModel Size=2.3B, Inference Mode=Offline2026.04 | 3.12 | |
| NIM4-ASRModel Size=2.3B, Inference Mode=Stream2026.04 | 3.28 | |
| OursModel Size=2.3B2026.04 | 3.45 | |
| Qwen3-ASR-1.7BModel Size=2.0B (↓)2026.04 | 4.22 | |
| Qwen3-ASR 1.7BModel Size=2.0B, Inference Mode=Offline2026.04 | 4.22 | |
| Step-Audio2-miniModel Size=8B+ (↑)2026.04 | 4.54 | |
| Step-Audio2 MiniModel Size=8B+, Inference Mode=Offline2026.04 | 4.54 | |
| Fun-ASR-nanoModel Size=0.8B (↓)2026.04 | 4.75 | |
| Fun-ASR NanoModel Size=0.8B, Inference Mode=Offline2026.04 | 4.75 | |
| Qwen3-Omni-InstModel Size=30B-A3B (↑)2026.04 | 5.19 | |
| Qwen3-Omni InstructModel Size=30B-A3B, Inference Mode=Offline2026.04 | 5.19 | |
| GLM-ASR-nanoModel Size=1.5B (↓)2026.04 | 7.25 | |
| GLM-ASR NanoModel Size=1.5B, Inference Mode=Offline2026.04 | 7.25 | |
| FireRedASR-LLMModel Size=8B+ (↑)2026.04 | 12.61 | |
| FireRedASR2S LLMModel Size=8B+, Inference Mode=Offline2026.04 | 13.4 |