Nanbeige releases Nanbeige4.3-33B model for coding and agents

Nanbeige has unveiled Nanbeige4.3-33B, a 33B-parameter flagship model for complex coding, tool-using agents and professional cowork tasks. It supports a 524,288-token context under Apache-2.0 and offers thinking and non-thinking modes. Vendor benchmarks show it ahead of Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B on most tests, though not all.

• 33B parameters, up to 524,288-token context, Apache-2.0 license
• Qwen3.8-27B scores higher on SWE-Bench Pro and GDPval, Gemma4-31B on GPQA Diamond
• Reasoning is toggled via enable_thinking and enabled by default
• XML is the recommended tool-call format, with JSON also supported