Ling-3.0 tiny & flash Base Models officially open-sourced
AI summary
Ant Group's Ling has open-sourced Ling-3.0-tiny-base and Ling-3.0-flash-base, releasing six checkpoints covering pretraining, mid-training and WSM merging. tiny-base has 7.9B total parameters with 1.3B activated, while flash-base has 124B total parameters with 5.1B activated. Both use a unified training recipe and suit continued pretraining, supervised fine-tuning, preference optimization and reinforcement learning post-training.
Key Takeaways
- Ant Group's Ling has open-sourced Ling-3.
- 0-tiny-base and Ling-3.
- 0-flash-base,releasing six checkpoints covering pretraining,mid-training and WSM merging.
- tiny-base has 7.
- 9B total parameters with 1.
Quick Facts
- Published
- Updated
- Source
- 公众号:蚂蚁百灵(Ling)
- Category
- Model Releases
- AI score
- 68
Ling-3.0 开放中间 checkpoint 和 WSM 合并策略,研究者可比较各训练阶段增益,并在自己数据上从合适节点继续预训练或后训练,而不只是拿到最终权重。
Source & copyright
This article is based on curated titles and AI summaries from the AIHOT public API, with attribution and links back. Original copyright remains with the original publishers and authors; this site does not mirror full texts.
How to Cite
When citing this article, keep the title, date, and permalink, and credit Orange News as the source.
Orange News: Ling-3.0 tiny & flash Base Models officially open-sourced. August 20, 2026. https://new.nat6.net/en/news/cmta7hu4g055groj2fuv1x60n/