GLM-5.2 is live. #1 throughput on OpenRouter, pay-per-token on FriendliAI. Try it today ➜
bosonai
Available on FriendliAI
Run this model inference on single tenant GPU with unmatched speed and reliability at scale.
Model Details
Model Provider
bosonai
Model Tree
Qwen/Qwen3-ASR-1.7B-hf
Input Modalities
Output Modalities
Supported Functionality
python
import torchfrom transformers import AutoProcessor, Qwen3ASRForConditionalGenerationmodel_id = "bosonai/Qwen3-ASR-1.7B-hf-orze"processor = AutoProcessor.from_pretrained(model_id)model = Qwen3ASRForConditionalGeneration.from_pretrained(model_id,dtype=torch.bfloat16,device_map="auto",)inputs = processor.apply_transcription_request(audio="audio.wav",language="en",).to(model.device, model.dtype)output_ids = model.generate(**inputs, max_new_tokens=512)generated_ids = output_ids[:, inputs["input_ids"].shape[1]:]text = processor.decode(generated_ids, return_format="transcription_only")[0]print(text)
The included generation configuration uses deterministic four-beam decoding with a 0.7 length penalty. This setting was selected using the Open ASR public test scores and public proxy evaluations; it changes decoding only and does not use private data.
The results below are self-evaluated public-set results, not maintainer-verified leaderboard results. The public test scores were used during checkpoint and decoding selection, so they are benchmark-fitted development results rather than an independent estimate of generalization. This model has no official private scripted or private conversational score or rank.
Neither the private Appen nor DataOcean leaderboard datasets was accessed. AppTek Call-Center Dialogues is a separate public dataset and is unrelated to Appen.
The results below were produced with the official leaderboard Transformers
evaluator and normalizer at commit
d1e99b25524814332d6868a5645e568670834cfb.
| Dataset | WER (%) |
|---|---|
| AMI Cleaned | 7.98 |
| Earnings22 Cleaned AA (chunked) | 5.78 |
| GigaSpeech Cleaned | 7.18 |
| LibriSpeech test-clean | 1.19 |
| LibriSpeech test-other | 2.75 |
| SPGISpeech | 2.70 |
| VoxPopuli Cleaned AA | 2.73 |
| Mean | 4.33 |
Measured throughput was 74.01 RTFx on the local A100 evaluation run. The same decoding settings were used for every dataset.
The final weights were compared at length penalties 0.8 and 0.7 on public English proxy sets used during development. These checks informed selection and are not independent validation. Lower WER is better.
| Proxy | lp=0.8 | lp=0.7 |
|---|---|---|
| AppTek multi-accent calls, selection 1 | 11.368 | 11.274 |
| AppTek multi-accent calls, selection 2 | 7.024 | 7.024 |
| AppTek multi-accent calls, selection 3 | 9.216 | 9.216 |
| MInDS-14 AU/GB/US holdout | 31.320 | 31.320 |
| Switchboard test, first 1,000 | 9.18 | 9.15 |
The three AppTek selections use different calls and were never used for training.
Orze first retained two compatible acoustic-side updates: an audio projector update at step 50 and an audio-tower linear-layer update at step 75. A final rank-16 audio-tower update was trained for 25 steps on a balanced public-data mix and fused at one-quarter strength. The final update's sampler drew from:
No Open ASR Leaderboard test audio, private Appen or DataOcean leaderboard data, selected AppTek proxy audio, or selected MInDS proxy partition was used for training. Candidate and decoding selection used the Open ASR public test scores and public proxy sets. The published checkpoint is fully merged and contains exactly 2,038,052,480 parameters in its safetensors metadata (2.04B when rounded to two decimals).
This specialization was optimized and evaluated for English short-form ASR. The multilingual abilities of the base model were not evaluated after tuning, so use the original Qwen checkpoint when broad multilingual performance is the priority. See the base model card for architecture, supported languages, safety considerations, and additional usage details.
Apache-2.0, following the base checkpoint.