K2 Horizon 0.9B needs about 1.6 GB of memory at Q4_K_M (a 0.7 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 5 quants, from 0.5 GB (Q3_K_M) to 1.1 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 1.9 GB.
K2 Horizon 0.9B – a compact dense decoder‑only model with 128K context
K2 Horizon 0.9B is a 1.08B‑parameter dense model in the Horizon family. It supports a 131072‑token context window and is trained with multi‑teacher distillation for math, code, STEM and instruction following. The model is fully open under the Apache‑2.0 license.
Good at
- Math reasoning
- Code generation
- Scientific QA
Limits
- No vision support
- Requires k2‑official llama.cpp
- Large context may need more VRAM
Best for
- Educational tutoring
- Research prototyping
- Tool‑integrated coding
Who it’s for. Developers and researchers who need a lightweight, high‑context language model for reasoning and coding tasks. The model is suitable for local deployment on modern GPUs with at least 8 GB VRAM.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q3_K_M | 0.5 GB | 1.5 GB | Fits | Get file → | |
| Q4_K_S | 0.6 GB | 1.6 GB | Fits | Get file → | |
| Q4_K_M | 0.7 GB | 1.6 GB | Fits | Get file → | |
| Q5_K_M | 0.8 GB | 1.7 GB | Fits | Get file → | |
| Q8_0 | 1.1 GB | 2.1 GB | Fits | Get file → |
For 8 GB VRAM, Q3_K_M (0.55 GB file, ~1.5 GB memory) or Q4_K_S (0.64 GB file, ~1.6 GB memory) are suitable. If 12 GB VRAM is available, Q4_K_M (0.67 GB file, ~1.6 GB memory) or Q5_K_M (0.77 GB file, ~1.7 GB memory) can be used. The Q8_0 quant (1.15 GB file, ~2.1 GB memory) requires at least 16 GB VRAM.
What GPU do you need for K2 Horizon 0.9B?
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 Q8_0 (1.1 GB file, about 2.1 GB in use).
Which K2 Horizon 0.9B quant should you choose?
Q4_K_M is the usual starting point for K2 Horizon 0.9B: it balances file size and quality well. Here it is a 0.7 GB file.
The Q8_0 file is 1.1 GB, about 1.7 times the size of Q4_K_M. It keeps the weights closer to the original, which only matters if you have the memory to spare.
If memory is tight, the smallest file is Q3_K_M at 0.5 GB, with the quality trade-offs that come with fewer bits. A reasonable choice when a 4-bit file will not fit, mostly on larger models. Quality loss is noticeable but often acceptable.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about K2 Horizon 0.9B
How much VRAM does K2 Horizon 0.9B need?
About 1.6 GB at Q4_K_M with an 8K context window. That is the 0.7 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 K2 Horizon 0.9B on an 8 GB GPU?
Yes, with Q8_0: the 1.1 GB file needs about 2.1 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of K2 Horizon 0.9B GGUF?
It depends on the quant: Q3_K_M is 0.5 GB, Q4_K_S is 0.6 GB, Q4_K_M is 0.7 GB, Q5_K_M is 0.8 GB, Q8_0 is 1.1 GB.
What context length does K2 Horizon 0.9B support?
The GGUF header lists a trained context length of 131,072 tokens (128K). The KV cache grows with the context you use: about 0.5 GB at 8K and 1.9 GB at 32K.
Where can I download K2 Horizon 0.9B GGUF?
The files are on Hugging Face in the darioooooo0o/K2-Horizon-0.9B-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 K2 Horizon 0.9B 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 K2 Horizon 0.9B on an RTX 3060 12GB GPU?
Yes, the model fits entirely on an RTX 3060 12GB. Use the Q4_K_M or Q5_K_M quant for optimal performance.
Does the model support vision inputs?
No, K2 Horizon 0.9B does not support vision inputs.
What llama.cpp branch is required?
Use the k2‑official branch of MBZUAI‑IFM/llama.cpp; mainline llama.cpp does not support the k2‑horizon architecture.