inferRouter/qwen3-asr-cs-1.7b

Qwen3-ASR 1.7B - Czech Q1c Fine-Tune

Offline Czech ASR fine-tuned from Qwen/Qwen3-ASR-1.7B-hf. This is the maximum-accuracy offline model in the InferRouter Czech ASR portfolio.

For live / streaming Czech ASR, use inferRouter/nemotron-cs-asr-0.6b.

This repository contains a merged standalone model: no PEFT adapter loading is required.

Main Result

This model is the accuracy tier. It beats Whisper-large-v3 on Czech formal/parliamentary speech and VoxPopuli in our evaluation matrix, while Whisper remains slightly ahead on FLEURS and Common Voice reference numbers.

WER is word-level, lowercased, punctuation-insensitive, and computed with the same frozen Czech evaluation board where locally measured. Whisper-large-v3 formal was measured locally on the same ParCzech formal split; the Whisper FLEURS/VoxPopuli/Common Voice values are included as external reference values.

model	mode	ParCzech formal	FLEURS	VoxPopuli	Common Voice
openai/whisper-large-v3	offline seq2seq reference	5.76	12.10	13.70	11.70
inferRouter/qwen3-asr-cs-1.7b	offline AED	3.43	12.37	9.79	11.76
inferRouter/nemotron-cs-asr-0.6b	streaming RNNT	5.87	17.18	11.28	14.28
internal Qwen3-ASR 0.6B full FT	offline AED	11.84	24.37	16.34	21.32

Interpretation

• Qwen 1.7B is the best model in this portfolio for offline / batch Czech ASR.
• It clearly beats Whisper-large-v3 on ParCzech formal and VoxPopuli.
• It is near Whisper-large-v3 on FLEURS/Common Voice but does not robustly beat Whisper there.
• Nemotron v4 is less accurate but is the true streaming / low-latency tier.
• The internal Qwen3-ASR 0.6B full fine-tune failed as a challenger and is not published.

Portfolio Position

model	best use	strength	tradeoff
inferRouter/qwen3-asr-cs-1.7b	offline / batch Czech ASR	best Czech formal/legal-adjacent accuracy, strong formatting	autoregressive AED; not true streaming
inferRouter/nemotron-cs-asr-0.6b	live / streaming Czech ASR	RNNT streaming, RTFx ~164, lower latency	lower WER ceiling
openai/whisper-large-v3	external offline baseline / teacher	strong multilingual general ASR	not the InferRouter streaming tier

Speed / Deployment Notes

• Batch-1 decode on RTX PRO 6000 Blackwell: RTFx ~3.1, mean latency around 2-3.5 s on measured short/medium clips.
• Batched offline throughput is much higher: dynamic batched eval reached roughly RTFx ~130 on the formal board.
• This is not a true streaming model. It is intended for record-then-transcribe or offline/batch transcription.

Training

• Base: Qwen/Qwen3-ASR-1.7B-hf.
• Q1b warm-start: intermediate Czech LoRA adaptation.
• Q1c: Czech full-anchor adaptation.
• Selected checkpoint: Q1c ckpt-85000 (best formal + low slot error among floor checkpoints), merged into the base model.

Detailed training-corpus composition is intentionally not listed in this public model card. Public benchmark test splits were held out from training.

Usage

import torch
from transformers import AutoProcessor, Qwen3ASRForConditionalGeneration

model_id = "inferRouter/qwen3-asr-cs-1.7b"
processor = AutoProcessor.from_pretrained(model_id)
model = Qwen3ASRForConditionalGeneration.from_pretrained(
    model_id,
    torch_dtype=torch.bfloat16,
    device_map="cuda",
)

# Use the Qwen3-ASR processor transcription template; set language="Czech" / "cs".

Intended Use

Offline Czech ASR for high-accuracy transcription, especially formal/parliamentary/legal-adjacent speech. For live captioning or low-latency streaming, use the Nemotron RNNT model instead.

Licence Notes

Base model is Apache-2.0. The merged model is released under Apache-2.0. Review the model licence and your downstream use case before redistribution or commercial deployment.