Model library / Nanbeige
Nanbeige4.3-33B is a 33.1 billion-parameter Llama-style causal model released by Nanbeige for high-capacity agentic tasks and teacher-model distillation.
Parameters in total
Context length in tokens, as published
Weights at the released precision
Licence declared on the repository
Overview
Nanbeige4.3-33B is a 33.1 billion-parameter Llama-style causal language model released by Nanbeige. The publisher describes it as a flagship teacher model for the Nanbeige4.3 series, intended to provide high-capacity agentic performance for coding, professional cowork tasks and to support downstream research and distillation into smaller models.
In deployment the model occupies roughly 66 GB in bfloat16 weight format and can process up to 524,288 tokens per request, enabling very long contexts. It is distributed under the Apache-2.0 licence, allowing commercial use, and can be run with Transformers, SGLang or vLLM inference frameworks.
Specification
| Repository | Nanbeige/Nanbeige4.3-33B |
| Publisher | Nanbeige |
| Published | 24 October 2026 |
| Architecture | LlamaForCausalLM |
| Model type | llama |
| Parameters | 33,118,977,024 |
| Layers | 52 |
| Hidden size | 7,168 |
| Attention heads | 56, 8 key/value heads |
| Context length | 524,288 tokens |
| Vocabulary | 166,912 tokens |
| Weight format | bfloat16 |
| Licence | apache-2.0 |
| Base model | Nanbeige/Nanbeige4.3-33B-Base |
Memory
Parameter count multiplied by the bytes each format uses. The key-value cache comes on top.
| Precision | Weights | Notes |
|---|---|---|
| bf16 | 61.7 GiB | Full precision as released |
| fp8 | 30.8 GiB | 8-bit, near-lossless on most models |
| int4 | 15.4 GiB | 4-bit, smallest footprint |
The cache costs 208 KiB per token at 16-bit, so the full 524,288-token context of one request needs about 104.0 GiB. Concurrency multiplies that number, not the weights.
Hardware
At the released precision, with 90% of the card given to the server and 2.5 GiB kept for the runtime. The context column is what is left for the cache on that setup.
| GPU | Cards needed | Free for cache | Context it holds |
|---|---|---|---|
| RTX 4090 24GB | 4 | 14.7 GiB | about 74,162 tokens |
| RTX 5090 32GB | 4 | 43.5 GiB | about 219,349 tokens |
| L40S 48GB | 2 | 19.7 GiB | about 99,368 tokens |
| A100 80GB | 1 | 7.8 GiB | about 39,377 tokens |
| H100 80GB | 1 | 7.8 GiB | about 39,377 tokens |
| RTX PRO 6000 96GB | 1 | 22.2 GiB | about 111,971 tokens |
| H200 141GB | 1 | 62.7 GiB | about 316,141 tokens |
Serving
A vLLM launch line for the shape above. Check the model card for a runtime the publisher recommends.
vllm serve Nanbeige/Nanbeige4.3-33B \ --max-model-len 73728 \ --gpu-memory-utilization 0.90
Questions
The weights take about 62 GiB at the released precision. That fits on one A100 80GB, which leaves roughly 8 GiB for the key-value cache.
The repository declares apache-2.0. Read the licence text before commercial use: the name alone does not tell you what is allowed.
The configuration allows 524,288 tokens. Whether the whole window is usable depends on the memory left for the cache, which the hardware table works out per card.
This page is a reference, and Aiqre does not serve this model today. We deploy open-weight models on dedicated EU hardware on request, so if you want an OpenAI-compatible endpoint for it, email [email protected] with the model name and roughly what volume you expect.
Also in the library
Nanbeige4.3-33B-Base is a 33.1-billion-parameter Llama-style causal language model released by Nanbeige under Apache-2.0.
Nanbeige4.2-3B is a compact agentic language model designed for reasoning, tool use and personal-assistant tasks.
Nanbeige4.2-3B is a 4.2 billion-parameter causal language model aimed at compact agentic and reasoning tasks.
Flint is a 39.8 billion-parameter language model for coding agents, using about 2.5 billion active parameters per token.
Hosting
Tell us the model and roughly what volume you expect. We reply with a price and an OpenAI-compatible endpoint, with your prompts never stored.