2 months ago
- Nanbeige4.2-3B is a compact agentic model with 3B non-embedding parameters, using a Looped Transformer architecture to increase capacity without adding parameters.
- It outperforms larger models like Qwen3.5-9B and Gemma4-12B on general-agent, code-agent, and reasoning benchmarks.
- The model is suitable for local deployment as a personal assistant, excelling in daily, office, and deep research tasks.
- Quickstart guides are provided for multiple frameworks including Transformers, SGLang, vLLM, llama.cpp, and Ollama.
- It supports configurable chat templates for reasoning and tool-use scenarios, with adjustable inference settings per task.
- Limitations include potential generation of inappropriate or biased outputs due to its probabilistic nature.