Biolm-30B

Biolm-30B (originally Qwen3-Omni-30B-A3B-Instruct) is a 30-billion-parameter (3B active) mixture-of-experts multimodal large language model.

About

Biolm-30B is designed for research on cross-modal temporal grounding and audio-visual understanding. It uses a Thinker-Talker MoE architecture to process text, images, audio, and video as input, and generates both text and natural speech as output. It supports 119 text languages, 19 speech input languages, and 10 speech output languages.

Modalities

Modality	Input	Output
Text	Yes	Yes
Image	Yes	—
Audio	Yes	Yes
Video	Yes	—

Intended Use

This model is intended for academic research on multimodal temporal reasoning, cross-modal attention analysis, and audio-visual language modeling.