Ling-1T
Ling-1T is the non-thinking flagship of Ant Group’s Ling 2.0 language-model series. A trillion-parameter MoE model activates approximately 50B parameters and uses evolutionary chain-of-thought training to improve efficient reasoning without long visible thinking traces. The family combines code generation, writing, instruction following, and tool calling with public MIT weights. Its original first-party API was retired in June 2026 in favor of Ling2.6; downloadable named checkpoints preserve historical access. The October 9 product changelog establishes public availability independently of earlier private repository uploads.
2025-10-09
1T total, approximately 50B active
Mixture of Experts Transformer with grouped-query attention
MIT
Specifications
- Parameters
- 1T total, approximately 50B active
- Architecture
- Mixture of Experts Transformer with grouped-query attention
- License
- MIT
- Context Window
- 131,072 tokens
- Type
- text
- Modalities
- text
Benchmark Scores
Advanced Specifications
- Model Family
- Ling
- API Access
- Not Available
- Chat Interface
- Not Available
- Variants
- Ling-1T-base
Capabilities & Limitations
- Capabilities
- reasoningcodingtool useinstruction following
- Known Limitations
- Inactive experts still require substantial weight storage.Tool use requires an external execution environment and appropriate chat template.
- Notable Use Cases
- coding agentsenterprise assistantslong-document analysis