bartowski/kai-os_Carnice-V3-GGUF

Llamacpp imatrix Quantizations of Carnice-V3 by kai-os

Using <a href="https://github.com/ggml-org/llama.cpp/">llama.cpp</a> release <a href="https://github.com/ggml-org/llama.cpp/releases/tag/b10603">b10603</a> for quantization.

Original model: https://huggingface.co/kai-os/Carnice-V3

Model details:

• —Parameter count: 28B
• —Input support: text, image (with mmproj file) - details
• —Speculative decoding: yes (MTP) - details
• —imatrix: yes - details

How to run

Prompt format

<|im_start|>system
Reasoning effort is set to xhigh. Please think carefully through the task, validate key assumptions, consider plausible alternatives, and prioritize correctness, consistency, and clarity in the final answer.

{system_prompt}<|im_end|>
<|im_start|>user
{prompt}<|im_end|>
<|im_start|>assistant
<think>

Don't know which to choose? Grab Q4_K_M (17.77GB) - usually a good mix of size and performance. Download instructions available here

Available files:

Download a specific file:

hf download bartowski/kai-os_Carnice-V3-GGUF --include "kai-os_Carnice-V3-Q4_K_M.gguf" --local-dir ./

Downloading using the Hugging Face CLI

<details> <summary>Click to view download instructions</summary>

First, make sure you have the Hugging Face CLI installed:

pip install -U "huggingface_hub[cli]"

Download a specific file:

hf download bartowski/kai-os_Carnice-V3-GGUF --include "kai-os_Carnice-V3-Q4_K_M.gguf" --local-dir ./

The files marked true in the Split column above are stored as multiple parts in a folder. To download all the parts to a local folder, run:

hf download bartowski/kai-os_Carnice-V3-GGUF --include "kai-os_Carnice-V3-bf16/*" --local-dir ./

You can either specify a new local-dir (kai-os_Carnice-V3-bf16) or download them all in place (./)

</details>

How to run

These quants run with llama.cpp - installable in one line via llama.app:

curl -LsSf https://llama.app/install.sh | sh
llama-server -hf bartowski/kai-os_Carnice-V3-GGUF:Q4_K_M

llama-server includes a built-in chat web UI, served at http://localhost:8080 by default.

These quants were made with llama.cpp release b10603 - if this model's architecture is newly supported, you'll need that release or newer to run them.

They also work in: LM Studio · koboldcpp · ramalama · Jan AI · Text Generation Web UI · LoLLMs · Atomic Chat

Multimodal

This model supports image input. Alongside the quants, this repo includes the multimodal projector files mmproj-kai-os_Carnice-V3-f16.gguf and mmproj-kai-os_Carnice-V3-bf16.gguf, which pair with any quant above.

llama.cpp downloads the mmproj automatically when using -hf as shown above; if you're loading files manually, pass it with --mmproj.

MTP

This model has MTP (Multi-Token Prediction) layers, and they are included in these quants

MTP layers act as a built-in draft model, letting llama.cpp run speculative decoding for faster generation. To use them, add the following flag to your llama.cpp command:

--spec-type draft-mtp

Note: the MTP layers are stored at Q40 in the imatrix quants (except for the Q80 quant), since imatrix calibration does not exercise them. Q4_0 is chosen for its speed which massively benefits MTP performance.

imatrix

All quants made using imatrix option, with a calibration corpus rendered through this model's own chat template. The corpus pairs plain prose with tool-calling and reasoning conversations (corpus source data), encoded exactly as this model sees them at inference and processed with --parse-special, so chat-format special tokens contribute to the importance matrix. The corpus rendered for this model is included in this repo: kai-os_Carnice-V3-calibration-v6.txt. The imatrix is available here: kai-os_Carnice-V3-imatrix.gguf.

<details> <summary>Calibration render details</summary>

{
  "generator": "auto_quant_v2 calibration renderer",
  "recipe": "calibration-v6",
  "model": "Carnice-V3",
  "encoder": "chat_template",
  "chunk_size": 512,
  "prose_chunks": 214,
  "tool_chunks": 369,
  "total_chunks": 583,
  "tool_chunk_fraction": 0.633,
  "n_conversations": 137,
  "extension_convs_used": 0,
  "conversation_token_lengths": [
    604,
    1667,
    1297,
    1559,
    1185,
    1404,
    3220,
    866,
    1267,
    1462,
    1107,
    2136,
    921,
    1293,
    2829,
    1313,
    1160,
    1026,
    777,
    758,
    1402,
    1098,
    1447,
    1242,
    1912,
    1538,
    1694,
    948,
    1455,
    1686,
    1609,
    1260,
    1308,
    1077,
    1060,
    1752,
    1691,
    1215,
    517,
    1950,
    1451,
    1166,
    1438,
    2042,
    2130,
    1355,
    1649,
    947,
    2996,
    1142,
    2908,
    827,
    1064,
    1010,
    947,
    738,
    2530,
    955,
    1188,
    1127,
    1271,
    1211,
    966,
    1259,
    1223,
    1628,
    953,
    1592,
    2166,
    881,
    365,
    1162,
    3375,
    2876,
    749,
    982,
    1072,
    1113,
    1338,
    1122,
    1191,
    835,
    1250,
    1111,
    1331,
    1586,
    1443,
    2106,
    913,
    698,
    2806,
    682,
    1443,
    1715,
    1987,
    1243,
    683,
    1396,
    1168,
    1769,
    1884,
    1755,
    857,
    1055,
    1077,
    2896,
    780,
    774,
    803,
    1452,
    1092,
    1617,
    808,
    392,
    361,
    2666,
    1058,
    1179,
    1870,
    2079,
    2735,
    2717,
    856,
    1020,
    880,
    999,
    1279,
    1039,
    882,
    1397,
    870,
    750,
    1783,
    1067,
    957,
    1356,
    1519
  ],
  "warnings": []
}

</details>

Embed/output weights

Some of these quants (Q3KXL, Q4KL etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

ARM/AVX information

llama.cpp automatically "repacks" weights into an interleaved layout at load time for faster inference on ARM and AVX machines - details in this PR. This once required downloading special Q4044/48/88 files; those are long gone. Online repacking now covers Q40, IQ4_NL, and most K-quants, so no special quant choice is needed for CPU inference.

Which file should I choose?

<details> <summary>Click here for details</summary>

An older (early 2024) but still useful write-up with charts comparing quant performances is provided by Artefact2 here

The first thing to figure out is how big a model you can run. To do this, you'll need to figure out how much RAM and/or VRAM you have.

If you want your model running as FAST as possible, you'll want to fit the whole thing on your GPU's VRAM. Aim for a quant with a file size 1-2GB smaller than your GPU's total VRAM.

If you want the absolute maximum quality, add both your system RAM and your GPU's VRAM together, then similarly grab a quant with a file size 1-2GB Smaller than that total.

Hugging Face can also do this math for you: add your hardware in your Local Apps settings and the model page will show which files fit.

Next, you'll need to decide if you want to use an 'I-quant' or a 'K-quant'.

If you don't want to think too much, grab one of the K-quants. These are in format 'QXKX', like Q5KM.

If you want to get more into the weeds, you can check out this extremely useful feature chart:

llama.cpp feature matrix

But basically, if you're aiming for below Q4, and you're running cuBLAS (Nvidia) or rocBLAS (AMD), you should look towards the I-quants. These are in format IQXX, like IQ3M. These are newer and offer better performance for their size.

These I-quants can also be used on CPU, but will be slower than their K-quant equivalent, so speed vs performance is a tradeoff you'll have to decide.

</details>

Credits

Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset.

Thank you ZeroWw for the inspiration to experiment with embed/output.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski