Nanbeige releases Nanbeige4.3-33B open agentic coding model

Nanbeige LLM Lab, part of recruiting platform BOSS Zhipin, published Nanbeige4.3-33B on Hugging Face under Apache 2.0 — a dense 33B model for agentic coding and office-style cowork tasks with a 524,288-token context window. The lab reports it leads 17 of 20 benchmarks against Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B. A Nanbeige4.3-33B-Base variant is also available for custom post-training.

• 33B parameters, dense architecture, 524,288-token context, BF16 weights
• Terminal-Bench 2.1: 80.9 versus 73.0 for Qwen3.8-27B
• SWE-Bench Pro: 60.3, trailing Qwen3.8-27B at 61.7
• BF16 weights need about 66 GB and multi-GPU servers