Automatic Speech Recognition
Transformers
Safetensors
Arabic
qwen3_asr
arabic
dialectal-arabic
qwen3-asr
Instructions to use oddadmix/qwen3-asr-0.6b-arabic-dialectal-v2 with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use oddadmix/qwen3-asr-0.6b-arabic-dialectal-v2 with Transformers:
# Use a pipeline as a high-level helper from transformers import pipeline pipe = pipeline("automatic-speech-recognition", model="oddadmix/qwen3-asr-0.6b-arabic-dialectal-v2")# Load model directly from transformers import AutoProcessor, AutoModelForMultimodalLM processor = AutoProcessor.from_pretrained("oddadmix/qwen3-asr-0.6b-arabic-dialectal-v2") model = AutoModelForMultimodalLM.from_pretrained("oddadmix/qwen3-asr-0.6b-arabic-dialectal-v2", device_map="auto") - Notebooks
- Google Colab
- Kaggle
comparison table: add cohere-...-dialectal-v2 (0.514)
Browse files
README.md
CHANGED
|
@@ -68,6 +68,7 @@ Same **dialect-balanced** test split (~2,600 clips, `oddadmix/lahgtna-v3-small`)
|
|
| 68 |
| whisper-small-v2 | 244M | 0.403 | 0.153 |
|
| 69 |
| nemotron-3.5-asr-v2 (streaming) | 638M | 0.423 | 0.148 |
|
| 70 |
| qwen3-asr-1.7b-v2 | 1.7B | 0.478 | 0.182 |
|
|
|
|
| 71 |
| qwen3-asr-0.6b-v2 | 0.6B | 0.551 | 0.215 |
|
| 72 |
|
| 73 |
`whisper-large-v3-turbo-v2` is the best; Whisper fine-tunes furthest. nemotron is the streaming option. (v1 models were trained on the earlier augmentation-expanded set.)
|
|
|
|
| 68 |
| whisper-small-v2 | 244M | 0.403 | 0.153 |
|
| 69 |
| nemotron-3.5-asr-v2 (streaming) | 638M | 0.423 | 0.148 |
|
| 70 |
| qwen3-asr-1.7b-v2 | 1.7B | 0.478 | 0.182 |
|
| 71 |
+
| cohere-transcribe-arabic-07-2026-dialectal-v2 | 2.07B | 0.514 | 0.294 |
|
| 72 |
| qwen3-asr-0.6b-v2 | 0.6B | 0.551 | 0.215 |
|
| 73 |
|
| 74 |
`whisper-large-v3-turbo-v2` is the best; Whisper fine-tunes furthest. nemotron is the streaming option. (v1 models were trained on the earlier augmentation-expanded set.)
|