GLM-5.3 is live. Run Z.ai's latest model on Friendli Model APIs. Try it today ➜
Kushtrim
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
Kushtrim
Model Tree
Qwen/Qwen3-ASR-1.7B
Input Modalities
Output Modalities
Supported Functionality
From the upstream model card, Qwen3-ASR provides:
qwen-asr toolkit.See the full upstream documentation here:
qwen-asr)Install:
bash
pip install -U qwen-asr
Load this model from a local folder (this directory) and transcribe:
python
import torchfrom qwen_asr import Qwen3ASRModelmodel = Qwen3ASRModel.from_pretrained("./",dtype=torch.bfloat16,device_map="cuda:0",max_inference_batch_size=32,max_new_tokens=256,)results = model.transcribe(audio="path/to/audio.wav",language="Norwegian", # or None for auto language ID)print(results[0].language)print(results[0].text)
Extracted from saved transformers training artifacts in this folder:
2e-5adamw_torch_fusedlanguage Norwegian<asr_text>... (see upstream docs); this fine-tune follows that convention.Qwen/Qwen3-ASR-1.7BThis work builds on Qwen3-ASR by the Qwen team. The original architecture and base weights remain the work of their respective authors. This repository documents the Norwegian fine-tuning and release work carried out by Kushtrim Visoka and does not claim authorship of the original Qwen3-ASR model.
If you use this fine-tuned model, please cite this repository and the original Qwen3-ASR model.
bibtex
@misc{visoka_qwen3_asr_17b_norwegian_2026,author = {Visoka, Kushtrim},title = {Qwen3-ASR-1.7B Norwegian},year = {2026},publisher = {Hugging Face},url = {https://huggingface.co/Kushtrim/Qwen3-ASR-1.7B-Norwegian},note = {Fine-tuned from Qwen/Qwen3-ASR-1.7B}}