Nanbeige4.3-33B-Base

Image: Nanbeige Logo

1. Introduction#

In Nanbeige4.3, we take a foundational step toward pushing the boundaries of lightweight agents. Our strategy centers on building a high-capacity teacher agent first, which will subsequently anchor the distillation of advanced agentic capabilities into smaller models.

To this end, we introduce Nanbeige4.3-33B, designed as the flagship teacher model for the Nanbeige4.3 series. Delivering state-of-the-art agentic performance among open models in the ~30B parameter class, Nanbeige4.3-33B excels at complex coding and professional cowork tasks. It not only offers a powerful, cost-effective option for immediate real-world deployment, but also lays the necessary foundation to power our upcoming small model distillation pipeline (e.g., for Nanbeige4.3-3B).

| Model | STEM Reasoning | Mathematics & Coding | General Knowledge

| MMLU-Pro | GPQA | SuperGPQA | GSM8K | Math | MBPP | MMLU | CMMLU

| Gemma4-31B-Base | 59.3 | 41.5 | 33.8 | 88.9 | 43.6 | 72.5 | 84.2 | 76.9

| Nanbeige4.3-33B-Base | 66.9 | 62.0 | 45.6 | 92.9 | 57.3 | 74.6 | 85.5 | 86.8

This repository contains Nanbeige4.3-33B-Base, the base model of Nanbeige4.3-33B. Compared with Gemma4-31B-Base, which is the only frontier open-source base model released at the 30B-parameter scale in the past six months, Nanbeige4.3-33B-Base consistently wins across a broad range of benchmarks. For usecase, Nanbeige4.3-33B-Base provides a strong foundation for post-training, whether to develop general-purpose or domain-specialized agents.

2. Quickstart#

Transformers#

from transformers import AutoModelForCausalLM, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained(
  'Nanbeige/Nanbeige4.3-33B-Base',
  use_fast=False,
  trust_remote_code=True
)
model = AutoModelForCausalLM.from_pretrained(
  'Nanbeige/Nanbeige4.3-33B-Base',
  torch_dtype='auto',
  device_map='auto',
  trust_remote_code=True
)

prompt = "The capital of China is"
input_ids = tokenizer(prompt, return_tensors='pt').input_ids
output_ids = model.generate(input_ids.to(model.device))
resp = tokenizer.decode(output_ids[0][len(input_ids[0]):], skip_special_tokens=True)
print(resp)

SGLang#

MODEL_PATH=/path/to/your/Nanbeige4.3-33B-Base
python -m sglang.launch_server \
  --model-path ${MODEL_PATH} \
  --host 0.0.0.0 \
  --port 8000 \
  --tp-size 4 \
  --mem-fraction-static 0.8

vLLM#

MODEL_PATH=/path/to/your/Nanbeige4.3-33B-Base
vllm serve ${MODEL_PATH} \
  --host 0.0.0.0 \
  --port 8000 \
  --tensor-parallel-size 4 \
  --gpu-memory-utilization 0.8

3. Contact#

If you have any questions, please open an issue in this repository or contact us at nanbeige@kanzhun.com.

We also welcome collaborations to build domain-specific and industry-focused agents, including joint work on data and training.

4. Citation#

If you find our model useful or would like to use it in your own work, please cite this Hugging Face project:

@misc{nanbeige43base2026,
  title        = {Nanbeige4.3-33B-Base},
  author       = {Nanbeige Team},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/Nanbeige/Nanbeige4.3-33B-Base}}
}