by Nanbeige
33B dense open-weights LLM from Nanbeige, tuned for agentic coding and professional cowork tasks.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Nanbeige4.3-33B is a 33-billion-parameter text-generation model from Nanbeige, released as the flagship "teacher" model of the Nanbeige4.3 series. It is a fine-tune of Nanbeige4.3-33B-Base and uses a Llama-style architecture in the Transformers format. The vendor says it will be used to distil agentic capabilities into smaller models, such as a planned Nanbeige4.3-3B.
The model card reports results against Qwen3.8-27B, Muse-Glimmer-30B and Gemma4-31B, all run at their highest thinking effort with the same harness. Nanbeige4.3-33B leads on most agentic coding benchmarks (Terminal-Bench 2.1, DeepSWE, NL2Repo) and cowork benchmarks (Automation Bench, JobBench, Toolathlon-Verified). Qwen3.8-27B scores higher on SWE-Bench Pro and GDPval, and Gemma4-31B scores higher on GPQA Diamond. These figures come from the vendor and have not been independently verified here. Nanbeige4.3-33B was evaluated with preserve_thinking=true in the chat template.
Weights are downloadable from Hugging Face under Apache 2.0, and the card lists English and Chinese as supported languages. The weights are about 66 GB, so self-hosting needs substantial GPU memory unless the model is quantized. The model is text-only.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
33B dense open model that beats similarly sized models including Qwen 3.8 and Gemma 4 on agentic and knowledge-work benchmarks. 66.3GB; code and model released.
Closest neighbours by meaning, not just by category.
Model · Nanbeige
3B-parameter open model from Nanbeige using a looped transformer, aimed at agentic and reasoning tasks.
Model · Alibaba
Alibaba's open-weights 2.4T-parameter MoE language model (95B active) from the Qwen3.8 family, with a hosted Max version.
Model · Alibaba
Alibaba's Qwen 3.5 large open-weights language model (~122B parameters) from the Qwen family.
Model
Open-weights 35B-A3B agentic model post-trained from Qwen3.6-35B-A3B for search, coding and scientific research.
Model · Meituan
Meituan's open-weights 1.6T-parameter mixture-of-experts LLM (48B active), aimed at agentic coding and reasoning tasks.
Model · Tencent
Tencent's open-weights 295B Mixture-of-Experts model (21B active) for reasoning and agentic coding.