Ant Ling, for a better humanity

Ant Ling is an LLM developed by Ant Group. Efficient, Intelligent, and Secure.

Jun Zhou's Talk at ZGC

Tech Talk

From Scaling Law to deeper thinking: 1T & omni modal practice

Jun Zhou: Scaling AGI at SITS

Tech Talk

From Language to Reasoning: Ant Ling's AGI Exploration

ICLR 2026 Paper Sharing Session

Tech Talk

WSM: Decay-Free Learning Rate Scheduling via Checkpoint Merging

Model Family

High-performance models for text, multimodal, and coding

Ling

Language

General-purpose LLM built on MoE architecture. Designed to maximize parameter scale while minimizing activation costs, achieving flagship intelligence with optimal cost-performance. A comprehensive model spectrum ranging from lightweight to 1T parameter scales.

Ling-2.6-1TLing-2.6-flashLing-2.6-mini
Model details

Ring

Reasoning

Deep reasoning LLM leveraging Ling's 1T parameter MoE architecture. Designed for high parameter scale with low activation costs, delivering industry-leading open-source results across competitive mathematics, programming, and integrated reasoning benchmarks.

Ring-2.6-1TRing-2.5-1T
Model details

Ming

Multimodal

Omni LLM featuring a unified architecture for seamless understanding and generation across text, image, audio, and video. The model achieves state-of-the-art performance in visual reasoning, immersive speech, and image synthesis.

Ming-flash-omni-2.0
Model details

Why Ant Ling

From Compute to Context: Building the Foundation for Enterprise AI Applications.

Superior Intelligence Efficiency

Flagship intelligence at 1/4 the cost of mainstream reasoning models

Production Grade Quality

Reliable in complex workflows: Multi-step task execution that stays robust

Developer Friendly

Integration with mainstream Agent frameworks for smoother tool calling and task orchestration

Comprehensive Solutions

From full-file coding to real-time completion; a complete solution from 1T flagships to Flash LLM