---
license: apache-2.0
base_model: MediaTek-Research/Breeze-ASR-26
language: [zh, nan]
pipeline_tag: automatic-speech-recognition
tags: [whisper, onnx, sherpa-onnx, taiwanese-hokkien, taigi, mobile, wasm, android, ios, int8, speech-recognition]
---

# Breeze-ASR-26 — ONNX (sherpa-onnx, INT8)

The **only** build that runs on **Android, iOS, and in the browser (WASM)**. Encoder
and decoder are single ONNX files each. Slower on CPU (RTF 1.3) than CT2/GGML — use
this only where you need mobile or WASM; otherwise prefer CT2 (speed) or GGML (low RAM).

> Part of the **Breeze-ASR-26 edge family** — the same MediaTek model in every runtime, pick by your constraint:
>
> | Repo | Runtime | RSS | RTF (CPU 4-thread) | Best for |
> |---|---|---|---|---|
> | [Breeze-ASR-26-ct2](https://huggingface.co/weemed/Breeze-ASR-26-ct2) | CTranslate2 / faster-whisper | ~2.9 GB | **0.21** | servers, 8 GB+ hosts, GPU |
> | [Breeze-ASR-26-GGML](https://huggingface.co/weemed/Breeze-ASR-26-GGML) | whisper.cpp / MacWhisper | **1.85 GB** | 0.40 | 4 GB hosts, desktop apps |
> | [Breeze-ASR-26-ONNX](https://huggingface.co/weemed/Breeze-ASR-26-ONNX) | sherpa-onnx / onnxruntime | — | 1.3 | Android / iOS / WASM |
>
> All Apache-2.0, derived from [MediaTek-Research/Breeze-ASR-26](https://huggingface.co/MediaTek-Research/Breeze-ASR-26). Measured on real multi-speaker Mandarin meeting audio. Mandarin does not regress; Taigi is transcribed as Mandarin meaning (not verbatim Taigi characters).

## Usage

```python
import sherpa_onnx
rec = sherpa_onnx.OfflineRecognizer.from_whisper(
    encoder="breeze-asr-26-large-v2-encoder.int8.onnx",
    decoder="breeze-asr-26-large-v2-decoder.int8.onnx",
    tokens="breeze-asr-26-large-v2-tokens.txt",
    language="zh", task="transcribe", num_threads=4)
```