Ternary-Bonsai-2-2B needs about 1.4 GB of memory at its main file (a 0.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 3 quants, from 0.4 GB to 0.5 GB (Q2_0). Longer context needs more: at 32K tokens the KV cache alone adds about 1.6 GB.
Ternary-Bonsai-2-2B: 2B vision‑language model with ternary weights
Ternary-Bonsai-2-2B is a 2B‑parameter vision‑language model based on Qwen3.5‑2B. It supports 262,144‑token context and uses ternary transformer weights stored in three GGUF packings. The model is available in llama.cpp for CUDA, Metal, and CPU.
Good at
- Vision‑language tasks
- Large context inference
- Efficient inference with ternary weights
Limits
- No text‑only mode without image projector
- Requires custom llama.cpp build for PTQ1_0 and PQ2_0
- License is Apache‑2.0
Best for
- Multimodal prototyping
- Research on ternary quantization
- Low‑memory inference
Who it’s for. Developers and researchers needing a compact vision‑language model for prototyping or fine‑tuning. The model is suitable for environments where GPU memory is limited but a 262K‑token context is required.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 0.4 GB | 1.3 GB | Fits | Get file → | |
| — | 0.5 GB | 1.4 GB | Fits | Get file → | |
| Q2_0 | 0.5 GB | 1.4 GB | Fits | Get file → |
For 8 GB VRAM, the Q2_0 pack (~0.55 GB) uses ~1.4 GB at 8 k context, fitting comfortably. If memory is tighter, PTQ1_0 (~0.43 GB) uses ~1.3 GB. The unknown packs are larger and not recommended for low‑memory setups.
What GPU do you need for Ternary-Bonsai-2-2B?
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: the best quant that fits is Q2_0 (0.5 GB file, about 1.4 GB in use).
Which Ternary-Bonsai-2-2B quant should you choose?
If memory is tight, the smallest file is the first one in the table at 0.4 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 Ternary-Bonsai-2-2B
How much VRAM does Ternary-Bonsai-2-2B need?
About 1.4 GB at its main file with an 8K context window. That is the 0.5 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 Ternary-Bonsai-2-2B on an 8 GB GPU?
Yes, with Q2_0: the 0.5 GB file needs about 1.4 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Ternary-Bonsai-2-2B GGUF?
It depends on the quant: file is 0.4 GB, file is 0.5 GB, Q2_0 is 0.5 GB.
What context length does Ternary-Bonsai-2-2B support?
The GGUF header lists a trained context length of 262,144 tokens (256K). The KV cache grows with the context you use: about 0.4 GB at 8K and 1.6 GB at 32K.
Where can I download Ternary-Bonsai-2-2B GGUF?
The files are on Hugging Face in the prism-ml/Ternary-Bonsai-2-2B-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 Ternary-Bonsai-2-2B 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 Ternary‑Bonsai‑2‑2B on 8 GB VRAM?
Yes, the Q2_0 pack uses about 1.4 GB at 8 k context, leaving headroom for other processes.
Does the model support text only inference?
No, the vision projector is required; the language model alone is not functional without image input.
Which quantization should I use for CPU inference?
Use the Q2_0 pack; it is supported by the standard llama.cpp build and works on CPU.