Llama3Ternary_10b needs about 4.1 GB of memory at TQ1_0 (a 2.5 GB file) with an 8K context window, so it fits on a 6 GB GPU. This repository has 2 quants, from 2.5 GB (TQ1_0) to 2.9 GB (TQ2_0).
Llama3Ternary_10b – a 10B ternary quantized Llama 3 model for low‑RAM inference
Llama3Ternary_10b is a 9.64B parameter Llama model with an expanded FFN, quantized to ternary –1, 0, 1. It is designed for Microsoft BitNet.cpp kernels and is not compatible with stock llama.cpp. The model supports a 8192 token context window.
Good at
- Low‑RAM chat inference
- Ternary quantization for fast inference
- BitNet.cpp compatibility
Limits
- Requires BitNet.cpp, not llama.cpp
- No vision or multimodal support
- License is llama3 community license
Best for
- Home chatbots on 8‑GB VRAM
- Small‑scale domain adaptation with LoRa
- Fast inference on AVX2/AVX‑512 CPUs
Who it’s for. Developers and small teams needing a lightweight Llama 3 variant for local deployment. The model is suitable for users who can run BitNet.cpp and want to fine‑tune with LoRa for domain adaptation.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| TQ1_0 | 2.5 GB | 4.1 GB | Fits | Get file → | |
| TQ2_0 | 2.9 GB | 4.5 GB | Fits | Get file → |
For 8‑GB VRAM, use the TQ1_0 quant (file size 2.55 GB, estimated memory 4.1 GB at 8k context). If you have 12‑GB VRAM, TQ2_0 (2.95 GB, 4.5 GB memory) provides a slightly higher quality option. Both estimates are approximate.
What GPU do you need for Llama3Ternary_10b?
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: the best quant that fits is TQ2_0 (2.9 GB file, about 4.5 GB in use).
Which Llama3Ternary_10b quant should you choose?
TQ1_0 (2.5 GB) is the middle option in this repository. Lower bit counts mean smaller files and more quality loss.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Llama3Ternary_10b
How much VRAM does Llama3Ternary_10b need?
About 4.1 GB at TQ1_0 with an 8K context window. That is the 2.5 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 Llama3Ternary_10b on an 8 GB GPU?
Yes, with TQ2_0: the 2.9 GB file needs about 4.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Llama3Ternary_10b GGUF?
It depends on the quant: TQ1_0 is 2.5 GB, TQ2_0 is 2.9 GB.
What context length does Llama3Ternary_10b support?
The GGUF header lists a trained context length of 8,192 tokens (8K). The KV cache grows with the context you use: about 1.1 GB at 8K and 1.1 GB at 8K.
Where can I download Llama3Ternary_10b GGUF?
The files are on Hugging Face in the mradermacher/Llama3Ternary_10b-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 Llama3Ternary_10b use?
Hugging Face lists the license as llama3. 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 Llama3Ternary_10b on 8 GB VRAM?
Yes, the TQ1_0 quant uses about 4.1 GB of VRAM at 8k context, which fits within an 8 GB budget.
Is this model compatible with llama.cpp?
No, it requires Microsoft BitNet.cpp kernels and will not load in stock llama.cpp.
What license applies to this model?
The model is licensed under the llama3 community license, as provided by Meta Platforms, Inc.