QURSE browser demo weights

Exact inference files used by the QURSE Tongue Twisters Demo. No training data, recordings, credentials, or evaluation banks are included.

asr.gguf: Qwen3-ASR-0.6B speech recognition decoder.
mmproj.gguf: matching Qwen audio projection / encoder.
detector.onnx: QURSE FP32 stumble detector, lexical seed 29.

weights.json records byte sizes and SHA256 digests. The demo pins an immutable Hub revision and verifies every downloaded or cached file before inference. The complete download is 2,639,870,960 bytes. Recordings stay on the player's device.

Qwen components retain Apache-2.0; see LICENSE-QWEN-APACHE-2.0.txt. Original QURSE contributions by woollyone retain CC-BY-NC-2.0: https://creativecommons.org/licenses/by-nc/2.0/legalcode . See NOTICE.md for attribution. This package contains inference weights only.

Downloads last month: 56

GGUF

Model size

0.8B params

Architecture

qwen3vl

Hardware compatibility

We're not able to determine the quantization variants.

View all variants

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Model tree for woollyone/QURSE-browser-demo

Base model

Qwen/Qwen3-ASR-0.6B

Quantized

(81)

this model