Description
Nanbeige4.3-33B is a flagship teacher model in the Nanbeige4.3 series, designed to push the boundaries of lightweight agents. It is a high-capacity model that excels in complex coding and professional cowork tasks, delivering state-of-the-art agentic performance among open models in the ~30B parameter class. This model not only provides a powerful and cost-effective option for immediate real-world deployment but also lays the foundation for a distillation pipeline aimed at smaller models, such as Nanbeige4.3-3B.
Read More
The model's performance is benchmarked against other frontier models like Qwen3.8-27B, Muse-Glimmer-30B, and Gemma4-31B, showing superior results in various categories including agentic coding and cowork tasks. It leads in most benchmarks, demonstrating strong performance across a wide range of professions and industries. For instance, on JobBench, it outperforms the baseline in most of the 35 occupations evaluated.
Nanbeige4.3-33B supports a context length of up to 524,288 tokens and offers configurable chat templates for reasoning and tool-use scenarios. It is compatible with Transformers, SGLang, and vLLM, providing flexibility in deployment. The model's tokenizer allows for customization in reasoning and tool-use scenarios, with options to enable or preserve thinking and to pass tools using formats like XML or JSON.
For inquiries or collaborations, users are encouraged to open an issue in the repository or contact the team via email. The model is open for collaborations to build domain-specific and industry-focused agents, including joint work on data and training.
- CategoryAI Models & LLMs
- Pricing ModelOpen Source
- DeploymentCloud
- PlatformsWeb
Nanbeige4.3-33B Highlights
High-capacity teacher model
Complex coding tasks
Professional cowork tasks
State-of-the-art agentic performance
Cost-effective deployment
Supports distillation pipeline
Context length up to 524,288 tokens
Compatible with Transformers, SGLang, vLLM
Getting Started with Nanbeige4.3-33B
Access model page: Visit the Hugging Face repository
Authenticate: Ensure you have access rights
Configure: Set up the model with desired parameters
Send prompts: Use the model for your tasks
Fine-tune: Adjust settings for specific needs
Nanbeige4.3-33B's Use Cases
- Complex Coding
- Professional Cowork
- Real-World Deployment
- Agentic Performance
- Small Model Distillation
