88Z AI, 21 August — Ant Group's Bailin large model open-sources Ling-3.0-tiny-base and Ling-3.0-flash-base, releasing weights for six key training stages. The Tiny version focuses on code and low-cost research, while the Flash version balances code, reasoning, and long-context capabilities.
At a glance
- Open-sources two base models and weights for six key training nodes
- Tiny version suits low-cost research; Flash version excels in reasoning and long context
The update
The Ant Group Bailin large model team has officially open-sourced two base models: Ling-3.0-tiny-base and Ling-3.0-flash-base. This release includes not only the final models but also the weights for six key training nodes across pre-training, mid-training, and WSM merging stages, providing the community with more transparent research resources. Ling-3.0-tiny-base features 7.9 billion total parameters and 1.3 billion active parameters, designed specifically for code generation and low-cost post-training research. Ling-3.0-flash-base possesses 124 billion total parameters and 5.1 billion active parameters, enhancing logical reasoning and long-context processing performance while maintaining code capabilities to meet more complex application needs.
Sources & editorial note
This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
- 百灵大模型Source published
Source dates above refer to the original announcements; the date under the headline is this article’s publication on 88Z AI.