Nanbeige LLM Lab has released Nanbeige4.3-33B, a 33 billion parameter dense language model, on Hugging Face under the Apache 2.0 license.
Key points
- Nanbeige4.3-33B is a 33B parameter open-source coding model.
- The model supports a 524,288-token context window.
- It leads 17 of 20 benchmarks against comparable models.
- Nanbeige4.3-33B is designed for agentic coding and 'cowork' tasks.
This model is built for agentic coding and office-style 'cowork' tasks. The lab, which is part of the recruitment platform BOSS Zhipin, also published Nanbeige4.3-33B-Base for custom post-training.
Nanbeige4.3-33B features a substantial 524,288-token context window. In internal benchmarks, the model reportedly leads 17 out of 20 benchmarks when compared against three other open models around the 30 billion parameter size, including Qwen3.8-27B, Muse-Glimmer-30B, and Gemma4-31B. Its best performance was 80.9 on Terminal-Bench 2.1, compared to Qwen3.8-27B's 73.0.
By the numbers
- parameter model size
- 33B
- token context window
- 524,288
- benchmarks led
- 17
The model's architecture is based on a standard LlamaForCausalLM configuration, with all 33 billion parameters active for every token processed. It includes 52 layers, a hidden size of 7,168, and uses grouped-query attention with 56 query heads and 8 key-value heads. The model supports a vocabulary of 166,912 tokens and a maximum of 524,288 positions.
Nanbeige4.3-33B's chat template includes an 'enable_thinking' control, which is on by default and toggles reasoning capabilities. The 'preserve_thinking' setting retains reasoning from previous turns. Nanbeige recommends the model for multi-turn tool use and code agents, and it supports tool calling via an XML call format.
Sources and references
Original report
Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model
MarkTechPost
Read on MarkTechPost (opens in a new tab)Linked in the original report
- Modelhuggingface.co/Nanbeige/Nanbeige4.3-33B (opens in a new tab)
- Paperarxiv.org/abs/2602.13367 (opens in a new tab)
- Modelhuggingface.co/Nanbeige/Nanbeige4.3-33B-Base (opens in a new tab)
- Postx.com/nanbeige/status/2114328933181333798 (opens in a new tab)
- Modelhuggingface.co/Nanbeige/Nanbeige4.3-33B/blob/main/conf… (opens in a new tab)
- Modelhuggingface.co/Qwen/Qwen3.8-27B (opens in a new tab)
Ref 00YJIACY
Cite this
APA
Interestana AI Editorial. (2026, October 25). Nanbeige Releases 33B Open Agentic Coding Model. Interestana. https://interestana.com/articles/nanbeige-releases-nanbeige43-33b-a-33b-open-agentic-coding-model-00yjiacy
MLA
“Nanbeige Releases 33B Open Agentic Coding Model.” Interestana, 25 Oct. 2026, interestana.com/articles/nanbeige-releases-nanbeige43-33b-a-33b-open-agentic-coding-model-00yjiacy.
Chicago
Interestana AI Editorial. “Nanbeige Releases 33B Open Agentic Coding Model.” Interestana, October 25, 2026. https://interestana.com/articles/nanbeige-releases-nanbeige43-33b-a-33b-open-agentic-coding-model-00yjiacy.
BibTeX
@misc{interestana_00yjiacy,
author = {{Interestana AI Editorial}},
title = {{Nanbeige Releases 33B Open Agentic Coding Model}},
howpublished = {Interestana},
year = {2026},
month = oct,
url = {https://interestana.com/articles/nanbeige-releases-nanbeige43-33b-a-33b-open-agentic-coding-model-00yjiacy},
note = {Summary of reporting by MarkTechPost, https://www.marktechpost.com/2026/10/25/nanbeige-releases-nanbeige4-3-33b-open-agentic-coding-model/}
}Credit the source
Original reporting: MarkTechPost, “Nanbeige Releases Nanbeige4.3-33B, a 33B Open Agentic Coding Model”, https://www.marktechpost.com/2026/10/25/nanbeige-releases-nanbeige4-3-33b-open-agentic-coding-model/(opens in a new tab)
Follow this story
Get Artificial Intelligence in your inbox — new stories in 40 words or fewer, once a week.
Subscribe