Automatic Speech Recognition on ASCEND
10.22Error Rate (CER/WER)FireRedASR2S LLM
Evaluation results
| Method | Links | |
|---|---|---|
| FireRedASR2S LLMModel Size=8B+, Inference Mode=Offline2026.04 | 10.22 | |
| Qwen3-ASR-1.7BModel Size=2.0B (↓)2026.04 | 10.87 | |
| Qwen3-ASR 1.7BModel Size=2.0B, Inference Mode=Offline2026.04 | 10.87 | |
| FireRedASR-LLMModel Size=8B+ (↑)2026.04 | 11.25 | |
| NIM4-ASRModel Size=2.3B, Inference Mode=Offline2026.04 | 11.46 | |
| OursModel Size=2.3B2026.04 | 11.79 | |
| NIM4-ASRModel Size=2.3B, Inference Mode=Stream2026.04 | 11.85 | |
| Fun-ASR-nanoModel Size=0.8B (↓)2026.04 | 11.91 | |
| Fun-ASR NanoModel Size=0.8B, Inference Mode=Offline2026.04 | 11.91 | |
| GLM-ASR-nanoModel Size=1.5B (↓)2026.04 | 12.29 | |
| GLM-ASR NanoModel Size=1.5B, Inference Mode=Offline2026.04 | 12.29 | |
| Step-Audio2-miniModel Size=8B+ (↑)2026.04 | 13.5 | |
| Step-Audio2 MiniModel Size=8B+, Inference Mode=Offline2026.04 | 13.5 | |
| Qwen3-Omni-InstModel Size=30B-A3B (↑)2026.04 | 18.68 | |
| Qwen3-Omni InstructModel Size=30B-A3B, Inference Mode=Offline2026.04 | 18.68 |