MoonshotAI
Featured
Kimi-K3
Kimi K3 is an open-weight, 2.8T-parameter native multimodal agentic model with a 1M-token context window, designed for frontier coding, knowledge work, and reasoning tasks.
LongCat-2.0
LongCat-2.0 is a 1.6-trillion-parameter MoE language model with ~48B activated per token, trained on AI ASIC superpods and featuring novel sparse attention and n-gram embedding for strong coding and agentic performance.
Stars
506
7-day growth
No data
Forks
57
Open issues
10
License
MIT
Last updated
2026-07-08
AI repository intelligence
FR-AI / ANALYSISWhy it is worth attention
It demonstrates frontier-scale training on alternative hardware (AI ASICs) without training instabilities, while introducing LongCat Sparse Attention and n-gram embedding to improve efficiency and long-context capability in a model that competes with leading proprietary systems.
Who it is for
- AI researchers studying large-scale MoE and sparse attention
- Developers building code agents and automated task workflows
- Organizations exploring training on non-GPU hardware
- Engineering teams needing long-context language models (1M tokens)
Use cases
- Repository-level code editing and understanding
- Automated task execution and agentic workflows
- Long-document analysis or reasoning over million-token contexts
- Integration with development harnesses like Claude Code and OpenClaw
Strengths
- Massive scale (1.6T total params) with efficient activation (48B) via MoE
- Novel LongCat Sparse Attention that reduces memory access overhead and indexing cost
- Competitive or leading scores on code and agent benchmarks (e.g., SWE-bench Pro 59.5, Terminal-Bench 2.1 70.8) compared to top proprietary models
- Trained on >35 trillion tokens with zero rollbacks on AI ASIC superpods, proving hardware stability
Considerations
- Not comprehensively evaluated for all downstream applications; performance may vary across languages and domains
- Large model size requires substantial inference resources (GPU or NPU) and specialized deployment configurations
- Limited public information on safety, fairness, and bias evaluations
README quick start
Usage Considerations
This model has not been specifically designed or comprehensively evaluated for every possible downstream application.
Related repositories
Similar projects matched by category, topics, and programming language.
xuchonglang
Featured
investing-for-beginners
A structured investing guide for Chinese beginners covering US stocks, options, and cryptocurrency, with focus on foundational concepts and risk awareness.
Krishnagangwal
Featured
CS-Fundamentals
A curated collection of Computer Science fundamentals (PDFs, notes, cheatsheets, interview question banks) for placement preparation, covering seven core subjects plus general resources.