Ming-Flash-Omni: A Sparse, Unified Architecture for Multimodal Perception and Generation

TL;DR

Ming-Flash-Omni is a highly efficient, large-scale multimodal model that unifies vision, speech, and language understanding and generation, advancing towards Artificial General Intelligence with improved performance and versatility.

Contribution

It introduces a sparse Mixture-of-Experts architecture with 100 billion parameters, enabling scalable, unified multimodal perception and generation across multiple modalities.

Findings

Achieves vision-language understanding comparable to Gemini 2.5 Pro

Excels in speech recognition and joint speech-sound-music generation

Demonstrates advanced generative semantic segmentation and in-image text rendering

Abstract

We propose Ming-Flash-Omni, an upgraded version of Ming-Omni, built upon a sparser Mixture-of-Experts (MoE) variant of Ling-Flash-2.0 with 100 billion total parameters, of which only 6.1 billion are active per token. This architecture enables highly efficient scaling (dramatically improving computational efficiency while significantly expanding model capacity) and empowers stronger unified multimodal intelligence across vision, speech, and language, representing a key step toward Artificial General Intelligence (AGI). Compared to its predecessor, the upgraded version exhibits substantial improvements across multimodal understanding and generation. Notably, it achieves strong performance on vision-language understanding benchmarks, with overall scores on par with Gemini 2.5 Pro, and enables seamless switching among multimodal tasks in multi-turn interactions. In speech, it achieves strong performance in contextual and dialect-aware ASR while enabling joint, continuous-generation of speech, sound, and music. In vision, it introduces generative semantic segmentation that achieves competitive standalone performance and enhances spatial control and editing consistency, alongside marked improvements in identity preservation, and high-fidelity in-image text rendering. Together, these capabilities demonstrate that a single unified model can serve as a practical foundation for general-purpose multimodal intelligence.