Carrot-Ko-2B-Instruct needs about 2.3 GB of memory at Q4_K_M (a 1.5 GB file) with an 8K context window, so it fits on a 4 GB GPU. This repository has 15 quants, from 1.0 GB (Q2_K) to 4.7 GB (F16). Longer context needs more: at 32K tokens the KV cache alone adds about 1.3 GB.
Carrot-Ko-2B-Instruct, a 2.34B parameter Qwen2-based instruct model
Carrot-Ko-2B-Instruct is a 2.34B parameter Qwen2 model fine‑tuned on Korean data. It supports 32768 context tokens and is released under the MIT license. The model is available in multiple GGUF quantizations.
Good at
- Korean language instruction
- Chat‑style Q&A
- Fine‑tuned Korean datasets
Limits
- No vision support
- Requires 8k context for memory estimates
- No explicit language list
Best for
- Chatbots in Korean
- Korean QA systems
- Local LLM deployment
Who it’s for. Developers needing a Korean‑focused instruct model with a moderate parameter count. The model is suitable for local deployment where GPU memory is limited.
Quants & memory
| Quant | File | Est. memory | vs VRAM | Result | |
|---|---|---|---|---|---|
| Q2_K | 1.0 GB | 1.8 GB | Fits | Get file → | |
| IQ3_XS | 1.1 GB | 1.9 GB | Fits | Get file → | |
| Q3_K_S | 1.1 GB | 1.9 GB | Fits | Get file → | |
| IQ3_S | 1.1 GB | 1.9 GB | Fits | Get file → | |
| IQ3_M | 1.1 GB | 2.0 GB | Fits | Get file → | |
| Q3_K_M | 1.2 GB | 2.0 GB | Fits | Get file → | |
| Q3_K_L | 1.3 GB | 2.1 GB | Fits | Get file → | |
| IQ4_XS | 1.3 GB | 2.2 GB | Fits | Get file → | |
| Q4_K_S | 1.4 GB | 2.2 GB | Fits | Get file → | |
| Q4_K_M | 1.5 GB | 2.3 GB | Fits | Get file → | |
| Q5_K_S | 1.6 GB | 2.5 GB | Fits | Get file → | |
| Q5_K_M | 1.7 GB | 2.5 GB | Fits | Get file → | |
| Q6_K | 1.9 GB | 2.8 GB | Fits | Get file → | |
| Q8_0 | 2.5 GB | 3.3 GB | Fits | Get file → | |
| F16 | 4.7 GB | 5.5 GB | Fits | Get file → |
For 8 GB VRAM, use Q4_K_M (file size 1.46 GB, memory 2.3 GB) or Q5_K_S (1.65 GB, 2.5 GB). For 4 GB VRAM, Q3_K_S (1.1 GB, 1.9 GB) or IQ3_XS (1.06 GB, 1.9 GB) are recommended. Higher precision quants like Q8_0 or F16 require more than 5 GB VRAM.
What GPU do you need for Carrot-Ko-2B-Instruct?
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 (2.5 GB file, about 3.3 GB in use).
- 6 GB VRAM: the best quant that fits is F16 (4.7 GB file, about 5.5 GB in use).
Which Carrot-Ko-2B-Instruct quant should you choose?
Q4_K_M is the usual starting point for Carrot-Ko-2B-Instruct: it balances file size and quality well. Here it is a 1.5 GB file.
The Q8_0 file is 2.5 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 Q2_K at 1.0 GB, with the quality trade-offs that come with fewer bits. Use it only when memory is very tight. Expect a clear drop in quality, which is usually worse on small models and easier to live with on large ones.
New to quant names? Read GGUF quantization explained.
Specifications
Architecture from the GGUF header
Source
Frequently asked questions about Carrot-Ko-2B-Instruct
How much VRAM does Carrot-Ko-2B-Instruct need?
About 2.3 GB at Q4_K_M with an 8K context window. That is the 1.5 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 Carrot-Ko-2B-Instruct on an 8 GB GPU?
Yes, with F16: the 4.7 GB file needs about 5.5 GB at 8K context. Larger quants in this repository do not fit in 8 GB.
What is the file size of Carrot-Ko-2B-Instruct GGUF?
It depends on the quant: Q2_K is 1.0 GB, IQ3_XS is 1.1 GB, Q3_K_S is 1.1 GB, IQ3_S is 1.1 GB, IQ3_M is 1.1 GB, Q3_K_M is 1.2 GB and 9 more.
What context length does Carrot-Ko-2B-Instruct support?
The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.3 GB at 8K and 1.3 GB at 32K.
Where can I download Carrot-Ko-2B-Instruct GGUF?
The files are on Hugging Face in the mradermacher/Carrot-Ko-2B-Instruct-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 Carrot-Ko-2B-Instruct use?
Hugging Face lists the license as mit. 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 Carrot-Ko-2B-Instruct on 8 GB VRAM?
Yes, using Q4_K_M or Q5_K_S quants will fit within 8 GB VRAM with approximate memory usage of 2.3–2.5 GB at 8k context.
Does the model support English?
The model was fine‑tuned on Korean data; English support is not guaranteed.
Is there a vision capability?
No, the model does not have vision support.