# Ant Ling — Ming-flash-omni-Preview: A Sparse, Unified Architecture for Multimodal Perception and Generation

- Company: Ant Ling (ant-ling.com)
- Announced: 2025-10-28
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://developer.ant-ling.com/en/blogs/ming-flash-omni-preview
- Record: https://forck.live/items/22335-ming-flash-omni-preview-a-sparse-unified-architecture-for-multimodal
- Subject: Ling / Ring / Ming
- Open weights: yes
- Models affected: Ming-flash-omni-Preview, Ming-lite-omni-1.5

Ant Ling released Ming-flash-omni-Preview, an open-source omnimodal large model with 103B total parameters and 9B activated, based on Ling 2.0's sparse MoE architecture. It improves over Ming-lite-omni-1.5 in omnimodal understanding and generation, achieving a GenEval score of 0.90 and SOTA on all 12 ContextASR subtasks. The model supports controllable image generation, streaming video understanding, speech and dialect recognition for 15 Chinese dialects, and voice cloning with continuous tokenizers.

## Evidence

Verbatim from https://developer.ant-ling.com/en/blogs/ming-flash-omni-preview:

> Ming-flash-omni-Preview is the first open-source omnimodal large model with a parameter scale reaching the hundred billion-Scale level. Based on Ling 2.0's sparse MoE architecture, Ming-flash-omni-Preview has a total of 103B parameters with 9B activated.

---

Record: https://forck.live/items/22335-ming-flash-omni-preview-a-sparse-unified-architecture-for-multimodal
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md