K2-Horizon-375B-A23B needs about 240.3 GB of memory at Q4_K_M (a 237.8 GB file) with an 8K context window, which is more than a 48 GB GPU holds. This repository has 25 quants, from 1.3 GB to 403.1 GB (Q8_0). Longer context needs more: at 32K tokens the KV cache alone adds about 8.2 GB.
K2-Horizon-375B-A23B: 375B sparse MoE model with 23B active params and 512K context
K2-Horizon-375B-A23B is a sparse Mixture-of-Experts model with 379B parameters and 23B active parameters per token. It supports a 524,288-token context window and is fully open source under the apache-2.0 license.
Good at
- Large context handling
- Sparse MoE efficiency
- Agentic tool use
Limits
- Requires high VRAM for full precision
- No vision support
- Large file size for full precision
Best for
- Long‑form document analysis
- Complex workflow automation
- Large‑scale text generation
Who it’s for. Developers and researchers needing a high‑capacity, open‑source LLM with extensive context. Suitable for applications that can allocate sufficient GPU memory for the chosen quantization.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| — | 1.3 GB | 3.8 GB | Fits | Get file → | |
| IQ1_S | 85.0 GB (3 parts) | 87.5 GB | Too big | Get file → | |
| IQ1_M | 94.7 GB (3 parts) | 97.3 GB | Too big | Get file → | |
| IQ2_XXS | 111.0 GB (3 parts) | 113.5 GB | Too big | Get file → | |
| IQ2_XS | 116.9 GB (3 parts) | 119.5 GB | Too big | Get file → | |
| IQ2_S | 124.0 GB (4 parts) | 126.6 GB | Too big | Get file → | |
| IQ2_M | 138.0 GB (4 parts) | 140.6 GB | Too big | Get file → | |
| Q2_K | 146.2 GB (4 parts) | 148.8 GB | Too big | Get file → | |
| IQ3_XXS | 161.9 GB (5 parts) | 164.5 GB | Too big | Get file → | |
| IQ3_XS | 172.1 GB (5 parts) | 174.6 GB | Too big | Get file → | |
| Q3_K_S | 173.1 GB (5 parts) | 175.6 GB | Too big | Get file → | |
| Q3_K_M | 182.5 GB (5 parts) | 185.1 GB | Too big | Get file → | |
| Q3_K_L | 192.3 GB (5 parts) | 194.9 GB | Too big | Get file → | |
| IQ3_M | 196.1 GB (5 parts) | 198.6 GB | Too big | Get file → | |
| IQ4_XS | 209.3 GB (6 parts) | 211.9 GB | Too big | Get file → | |
| Q4_0 | 216.1 GB (6 parts) | 218.7 GB | Too big | Get file → | |
| Q4_K_S | 224.0 GB (6 parts) | 226.6 GB | Too big | Get file → | |
| Q4_K_M | 237.8 GB (7 parts) | 240.3 GB | Too big | Get file → | |
| IQ4_NL | 237.8 GB (7 parts) | 240.4 GB | Too big | Get file → | |
| Q4_1 | 238.6 GB (7 parts) | 241.2 GB | Too big | Get file → | |
| Q5_K_S | 269.7 GB (7 parts) | 272.3 GB | Too big | Get file → | |
| Q5_K_M | 289.6 GB (8 parts) | 292.1 GB | Too big | Get file → | |
| Q6_K_S | 319.1 GB (9 parts) | 321.7 GB | Too big | Get file → | |
| Q6_K | 334.5 GB (9 parts) | 337.0 GB | Too big | Get file → | |
| Q8_0 | 403.1 GB (11 parts) | 405.6 GB | Too big | Get file → |
For 8 GB VRAM, use Q4_K_M (237.80 GB file, ~240.3 GB memory at 8k context). For 16 GB VRAM, Q4_K_S (224.01 GB file, ~226.6 GB memory) or Q4_0 (216.15 GB file, ~218.7 GB memory). For 32 GB VRAM, Q5_K_M (289.59 GB file, ~292.1 GB memory) or Q6_K (334.48 GB file, ~337 GB memory).
What GPU do you need for K2-Horizon-375B-A23B?
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 the main file (1.3 GB file, about 3.8 GB in use).
- 6 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
- 8 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
- 12 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
- 16 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
- 24 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
- 32 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
- 48 GB VRAM: the best quant that fits is the main file (1.3 GB file, about 3.8 GB in use).
Which K2-Horizon-375B-A23B quant should you choose?
Q4_K_M is the usual starting point for K2-Horizon-375B-A23B: it balances file size and quality well. Here it is a 237.8 GB file.
The Q8_0 file is 403.1 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 the first one in the table at 1.3 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 K2-Horizon-375B-A23B
How much VRAM does K2-Horizon-375B-A23B need?
About 240.3 GB at Q4_K_M with an 8K context window. That is the 237.8 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-375B-A23B on an 8 GB GPU?
Yes, with the smaller files: the 1.3 GB file needs about 3.8 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of K2-Horizon-375B-A23B GGUF?
It depends on the quant: file is 1.3 GB, IQ1_S is 85.0 GB, IQ1_M is 94.7 GB, IQ2_XXS is 111.0 GB, IQ2_XS is 116.9 GB, IQ2_S is 124.0 GB and 19 more.
What context length does K2-Horizon-375B-A23B support?
The GGUF header lists a trained context length of 524,288 tokens (512K). The KV cache grows with the context you use: about 2.0 GB at 8K and 8.2 GB at 32K.
Where can I download K2-Horizon-375B-A23B GGUF?
The files are on Hugging Face in the bartowski/K2-Horizon-375B-A23B-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-375B-A23B 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-375B-A23B on 8 GB VRAM?
Use Q4_K_M quantization; it requires about 240.3 GB memory at 8k context, so 8 GB VRAM is insufficient for full inference.
Does the model support vision inputs?
No, the model only supports text input.
What license applies to this model?
The model is released under the apache-2.0 license.