🤖 Nanbeige4.2-3B: A Powerful AI Agent for Local Execution
Nanbeige4.2-3B has been released—a compact agentic model with 3 billion non-embedding parameters based on the Looped Transformer architecture. Despite its small size, it outperforms Qwen3.5-9B and Gemma4-12B in agentic tasks and coding. The model supports a context window of up to 256k tokens and is optimized for llama.cpp and Ollama.
🌍 It demonstrates the efficiency of architectures with layer reuse, allowing for performance levels comparable to large models with a small number of parameters.
👤 It enables running a powerful AI agent for task automation directly on local hardware.