HEBATRON needs about 32.0 GB of memory at Q4_K_M (a 24.5 GB file) with an 8K context window, so it fits on a 32 GB GPU. This repository has 24 quants, from 0.1 GB to 33.5 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 27.9 GB.
Hebrew‑specialized Mamba2‑MoE language model
HEBATRON is a 31.6B‑parameter model built on the nemotron_h_moe architecture. It supports a 1,048,576‑token context window and is licensed under apache-2.0.
Good at
- Hebrew reasoning
- Long‑context summarization
- Mixed‑language support
Limits
- No vision support
- Requires large VRAM for high‑quality quants
- No explicit language list
Best for
- Hebrew document analysis
- Legal text summarization
- Academic research in Hebrew
Who it’s for. Developers building Hebrew‑centric NLP tools or researchers needing deep Hebrew language understanding. The model is suitable for applications that require extended context and native‑level reasoning.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.1 GB | 7.5 GB | Tight | Get file → | |
| IQ1_S | 17.9 GB | 25.3 GB | Too big | Get file → | |
| IQ1_M | 17.9 GB | 25.3 GB | Too big | Get file → | |
| IQ2_XXS | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ2_XS | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ2_S | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ2_M | 17.9 GB | 25.4 GB | Too big | Get file → | |
| Q2_K | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ3_XXS | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ3_S | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ3_XS | 17.9 GB | 25.4 GB | Too big | Get file → | |
| Q3_K_S | 17.9 GB | 25.4 GB | Too big | Get file → | |
| IQ4_XS | 18.0 GB | 25.4 GB | Too big | Get file → | |
| Q4_0 | 18.1 GB | 25.5 GB | Too big | Get file → | |
| Q2_K_S | 18.1 GB | 25.5 GB | Too big | Get file → | |
| IQ3_M | 18.1 GB | 25.6 GB | Too big | Get file → | |
| Q3_K_M | 19.8 GB | 27.3 GB | Too big | Get file → | |
| Q4_1 | 19.9 GB | 27.4 GB | Too big | Get file → | |
| Q3_K_L | 20.7 GB | 28.2 GB | Too big | Get file → | |
| Q4_K_S | 21.9 GB | 29.4 GB | Too big | Get file → | |
| Q5_K_S | 23.8 GB | 31.3 GB | Too big | Get file → | |
| Q4_K_M | 24.5 GB | 32.0 GB | Too big | Get file → | |
| Q5_K_M | 26.0 GB | 33.5 GB | Too big | Get file → | |
| Q6_K | 33.5 GB | 41.0 GB | Too big | Get file → |
For 8 GB VRAM, use IQ1_S or IQ1_M (file size 17.86 GB, estimated 25.3 GB memory). For 12 GB VRAM, Q2_K (file size 17.91 GB, estimated 25.4 GB memory). For 24 GB VRAM, Q4_K_M (file size 24.52 GB, estimated 32 GB memory).
What GPU do you need for HEBATRON?
These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.
- 4 GB VRAM: no quant fits at an 8K context.
- 6 GB VRAM: no quant fits at an 8K context.
- 8 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 7.5 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 7.5 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 7.5 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 7.5 GB in use).
- 32 GB VRAM: the best quant that fits is Q4_K_M (24.5 GB file, about 32.0 GB in use).
- 48 GB VRAM: the best quant that fits is Q6_K (33.5 GB file, about 41.0 GB in use).
Which HEBATRON quant should you choose?
Q4_K_M is the usual starting point for HEBATRON: it balances file size and quality well. Here it is a 24.5 GB file.
If memory is tight, the smallest file is the first one in the table at 0.1 GB, with the quality trade-offs that come with fewer bits.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about HEBATRON
How much VRAM does HEBATRON need?
About 32.0 GB at Q4_K_M with an 8K context window. That is the 24.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.
Can I run HEBATRON on an 8 GB GPU?
Yes, with the smaller files: the 0.1 GB file needs about 7.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of HEBATRON GGUF?
It depends on the quant: file is 0.1 GB, IQ1_S is 17.9 GB, IQ1_M is 17.9 GB, IQ2_XXS is 17.9 GB, IQ2_XS is 17.9 GB, IQ2_S is 17.9 GB and 18 more.
What context length does HEBATRON support?
The GGUF header lists a trained context length of 1,048,576 tokens (1024K). The KV cache grows with the context you use: about 7.0 GB at 8K and 27.9 GB at 32K.
Where can I download HEBATRON GGUF?
The files are on Hugging Face in the mradermacher/Hebatron_base-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.
What license does HEBATRON use?
Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.
Can I run HEBATRON on 8 GB VRAM?
Use the IQ1_S or IQ1_M quant, which estimate 25.3 GB memory at 8k context. 8 GB VRAM is insufficient for these quants.
Does HEBATRON support English?
Yes, it supports Hebrew and English, but the primary focus is Hebrew language tasks.