# Ant Ling — Ling-3.0-flash: More Useful Work per Token

- Company: Ant Ling (ant-ling.com)
- Announced: 2026-07-24
- Category: new-model
- Coverage: not counted
- Announcement: yes
- Group: models
- Source: https://developer.ant-ling.com/en/blogs/ling-3.0-flash-release
- Record: https://forck.live/items/22330-ling-3-0-flash-more-useful-work-per-token
- Subject: Ling / Ring / Ming
- Models affected: Ling-3.0-flash
- Context window: Native 256K context, extendable to 1M

Ant Ling released Ling-3.0-flash, a 124B-parameter Mixture-of-Experts model with 5.1B active parameters per token, designed for token efficiency in agent workflows. The model features native 256K context extendable to 1M, hybrid linear attention (KDA + MLA), and agent-focused training across over 10,000 environments. It achieves a 25.3% pass rate on MiniAppBench, competitive with open-source models roughly twice its size.

## Evidence

Verbatim from https://developer.ant-ling.com/en/blogs/ling-3.0-flash-release:

> Ling-3.0-flash is AntLing's attempt to push token efficiency into the model design itself. It is a 124B-parameter Mixture-of-Experts model with only 5.1B active parameters per token.

---

Record: https://forck.live/items/22330-ling-3-0-flash-more-useful-work-per-token
Catalogue: https://forck.live/llms.txt
Current issue: https://forck.live/feed.md