GraySoft
Projects Models Compare Cloud benchmarks FAQ Download guIDE →
Model Intelligence Sheet

bartowski/Ling-3.0-flash-GGUF overview

Llamacpp imatrix Quantizations of Ling 3.0 flash by inclusionAI Using <a href="https://github.com/ggml org/llama.cpp/" llama.cpp</a release <a href="https://gi…

gguftext-generationbase_model:inclusionAI/Ling-3.0-flashbase_model:quantized:inclusionAI/Ling-3.0-flashlicense:mitendpoints_compatibleregion:usimatrixconversational

Runs locally from ~465.2 MB disk (4 GB VRAM class GPUs with llama.cpp / guIDE).

Downloads
0
Likes
4
Pipeline
text-generation
Author

Repository Files & Downloads

61 GGUF files detected
Direct downloads for local inference
FileTypeQuantizationSizeLink
Ling-3.0-flash-IQ1_M.ggufGGUFIQ1_M28.46 GBDownload
Ling-3.0-flash-IQ1_S.ggufGGUFIQ1_S25.68 GBDownload
Ling-3.0-flash-IQ2_M.ggufGGUFIQ2_M40.64 GBDownload
Ling-3.0-flash-IQ2_S.ggufGGUFIQ2_S36.97 GBDownload
Ling-3.0-flash-IQ2_XS.ggufGGUFIQ2_XS36.36 GBDownload
Ling-3.0-flash-IQ2_XXS.ggufGGUFIQ2_XXS32.82 GBDownload
Ling-3.0-flash-IQ3_M/Ling-3.0-flash-IQ3_M-00001-of-00002.ggufGGUFIQ3_M37.01 GBDownload
Ling-3.0-flash-IQ3_M/Ling-3.0-flash-IQ3_M-00002-of-00002.ggufGGUFIQ3_M20.28 GBDownload
Ling-3.0-flash-IQ3_XS/Ling-3.0-flash-IQ3_XS-00001-of-00002.ggufGGUFIQ3_XS37.24 GBDownload
Ling-3.0-flash-IQ3_XS/Ling-3.0-flash-IQ3_XS-00002-of-00002.ggufGGUFIQ3_XS17.65 GBDownload
Ling-3.0-flash-IQ3_XXS/Ling-3.0-flash-IQ3_XXS-00001-of-00002.ggufGGUFIQ3_XXS37.14 GBDownload
Ling-3.0-flash-IQ3_XXS/Ling-3.0-flash-IQ3_XXS-00002-of-00002.ggufGGUFIQ3_XXS13.26 GBDownload
Ling-3.0-flash-IQ4_NL/Ling-3.0-flash-IQ4_NL-00001-of-00002.ggufGGUFIQ4_NL37.14 GBDownload
Ling-3.0-flash-IQ4_NL/Ling-3.0-flash-IQ4_NL-00002-of-00002.ggufGGUFIQ4_NL30.45 GBDownload
Ling-3.0-flash-IQ4_XS/Ling-3.0-flash-IQ4_XS-00001-of-00002.ggufGGUFIQ4_XS37.19 GBDownload
Ling-3.0-flash-IQ4_XS/Ling-3.0-flash-IQ4_XS-00002-of-00002.ggufGGUFIQ4_XS26.83 GBDownload
Ling-3.0-flash-Q2_K.ggufGGUFQ2_K42.75 GBDownload
Ling-3.0-flash-Q2_K_L.ggufGGUFQ2_K_L43.11 GBDownload
Ling-3.0-flash-Q3_K_L/Ling-3.0-flash-Q3_K_L-00001-of-00002.ggufGGUFQ3_K_L36.96 GBDownload
Ling-3.0-flash-Q3_K_L/Ling-3.0-flash-Q3_K_L-00002-of-00002.ggufGGUFQ3_K_L20.25 GBDownload
Ling-3.0-flash-Q3_K_M/Ling-3.0-flash-Q3_K_M-00001-of-00002.ggufGGUFQ3_K_M37.25 GBDownload
Ling-3.0-flash-Q3_K_M/Ling-3.0-flash-Q3_K_M-00002-of-00002.ggufGGUFQ3_K_M17.67 GBDownload
Ling-3.0-flash-Q3_K_S/Ling-3.0-flash-Q3_K_S-00001-of-00002.ggufGGUFQ3_K_S37.02 GBDownload
Ling-3.0-flash-Q3_K_S/Ling-3.0-flash-Q3_K_S-00002-of-00002.ggufGGUFQ3_K_S15.40 GBDownload
Ling-3.0-flash-Q3_K_XL/Ling-3.0-flash-Q3_K_XL-00001-of-00002.ggufGGUFQ3_K_XL36.87 GBDownload
Ling-3.0-flash-Q3_K_XL/Ling-3.0-flash-Q3_K_XL-00002-of-00002.ggufGGUFQ3_K_XL20.66 GBDownload
Ling-3.0-flash-Q4_0/Ling-3.0-flash-Q4_0-00001-of-00002.ggufGGUFQ4_036.88 GBDownload
Ling-3.0-flash-Q4_0/Ling-3.0-flash-Q4_0-00002-of-00002.ggufGGUFQ4_030.96 GBDownload
Ling-3.0-flash-Q4_1/Ling-3.0-flash-Q4_1-00001-of-00003.ggufGGUFQ4_136.94 GBDownload
Ling-3.0-flash-Q4_1/Ling-3.0-flash-Q4_1-00002-of-00003.ggufGGUFQ4_137.23 GBDownload
Ling-3.0-flash-Q4_1/Ling-3.0-flash-Q4_1-00003-of-00003.ggufGGUFQ4_1548.1 MBDownload
Ling-3.0-flash-Q4_K_L/Ling-3.0-flash-Q4_K_L-00001-of-00002.ggufGGUFQ4_K_L37.15 GBDownload
Ling-3.0-flash-Q4_K_L/Ling-3.0-flash-Q4_K_L-00002-of-00002.ggufGGUFQ4_K_L35.59 GBDownload
Ling-3.0-flash-Q4_K_M/Ling-3.0-flash-Q4_K_M-00001-of-00002.ggufGGUFQ4_K_M36.87 GBDownload
Ling-3.0-flash-Q4_K_M/Ling-3.0-flash-Q4_K_M-00002-of-00002.ggufGGUFQ4_K_M35.59 GBDownload
Ling-3.0-flash-Q4_K_S/Ling-3.0-flash-Q4_K_S-00001-of-00002.ggufGGUFQ4_K_S37.20 GBDownload
Ling-3.0-flash-Q4_K_S/Ling-3.0-flash-Q4_K_S-00002-of-00002.ggufGGUFQ4_K_S32.69 GBDownload
Ling-3.0-flash-Q5_K_L/Ling-3.0-flash-Q5_K_L-00001-of-00003.ggufGGUFQ5_K_L37.25 GBDownload
Ling-3.0-flash-Q5_K_L/Ling-3.0-flash-Q5_K_L-00002-of-00003.ggufGGUFQ5_K_L36.69 GBDownload
Ling-3.0-flash-Q5_K_L/Ling-3.0-flash-Q5_K_L-00003-of-00003.ggufGGUFQ5_K_L10.73 GBDownload
Ling-3.0-flash-Q5_K_M/Ling-3.0-flash-Q5_K_M-00001-of-00003.ggufGGUFQ5_K_M37.02 GBDownload
Ling-3.0-flash-Q5_K_M/Ling-3.0-flash-Q5_K_M-00002-of-00003.ggufGGUFQ5_K_M36.69 GBDownload
Ling-3.0-flash-Q5_K_M/Ling-3.0-flash-Q5_K_M-00003-of-00003.ggufGGUFQ5_K_M10.73 GBDownload
Ling-3.0-flash-Q5_K_S/Ling-3.0-flash-Q5_K_S-00001-of-00003.ggufGGUFQ5_K_S37.19 GBDownload
Ling-3.0-flash-Q5_K_S/Ling-3.0-flash-Q5_K_S-00002-of-00003.ggufGGUFQ5_K_S37.13 GBDownload
Ling-3.0-flash-Q5_K_S/Ling-3.0-flash-Q5_K_S-00003-of-00003.ggufGGUFQ5_K_S7.53 GBDownload
Ling-3.0-flash-Q6_K/Ling-3.0-flash-Q6_K-00001-of-00003.ggufGGUFQ6_K36.63 GBDownload
Ling-3.0-flash-Q6_K/Ling-3.0-flash-Q6_K-00002-of-00003.ggufGGUFQ6_K36.67 GBDownload
Ling-3.0-flash-Q6_K/Ling-3.0-flash-Q6_K-00003-of-00003.ggufGGUFQ6_K28.46 GBDownload
Ling-3.0-flash-Q8_0/Ling-3.0-flash-Q8_0-00001-of-00004.ggufGGUFQ8_036.68 GBDownload
Ling-3.0-flash-Q8_0/Ling-3.0-flash-Q8_0-00002-of-00004.ggufGGUFQ8_036.68 GBDownload
Ling-3.0-flash-Q8_0/Ling-3.0-flash-Q8_0-00003-of-00004.ggufGGUFQ8_036.68 GBDownload
Ling-3.0-flash-Q8_0/Ling-3.0-flash-Q8_0-00004-of-00004.ggufGGUFQ8_016.28 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00001-of-00007.ggufGGUFBF1636.95 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00002-of-00007.ggufGGUFBF1635.63 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00003-of-00007.ggufGGUFBF1635.63 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00004-of-00007.ggufGGUFBF1635.63 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00005-of-00007.ggufGGUFBF1635.63 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00006-of-00007.ggufGGUFBF1635.63 GBDownload
Ling-3.0-flash-bf16/Ling-3.0-flash-bf16-00007-of-00007.ggufGGUFBF1622.50 GBDownload
Ling-3.0-flash-imatrix.ggufGGUFGGUF465.2 MBDownload

Model Details

Model IDbartowski/Ling-3.0-flash-GGUF
Authorbartowski
Pipelinetext-generation
Licensemit
Base modelinclusionAI/Ling-3.0-flash
Last modified2026-08-18T15:26:21.000Z

Model README

---

quantized_by: bartowski

pipeline_tag: text-generation

license: mit

base_model: inclusionAI/Ling-3.0-flash

base_model_relation: quantized

---

Llamacpp imatrix Quantizations of Ling-3.0-flash by inclusionAI

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/b10472">b10472</a> for quantization.

Original model: https://huggingface.co/inclusionAI/Ling-3.0-flash

Model details:

  • Parameter count: 127B
  • Input support: text
  • Speculative decoding: yes (MTP) - details
  • imatrix: yes - details
  • Perplexity/KLD measured: no

How to run

Prompt format

<role>SYSTEM</role>{system_prompt}
detailed thinking on<|role_end|><role>HUMAN</role>{prompt}<|role_end|><role>ASSISTANT</role>
<think>

Don't know which to choose? Grab Q4_K_M (77.80GB) - usually a good mix of size and performance. Download instructions available here

Available files:

| Filename | Quant type | File Size | Split | Description |

| -------- | ---------- | --------- | ----- | ----------- |

| Ling-3.0-flash-bf16.gguf | bf16 | 255.09GB | true | Full BF16 weights. |

| Ling-3.0-flash-Q8_0.gguf | Q8_0 | 135.63GB | true | Extremely high quality, generally unneeded but max available quant. |

| Ling-3.0-flash-Q6_K.gguf | Q6_K | 109.26GB | true | Very high quality, near perfect, recommended. |

| Ling-3.0-flash-Q5_K_L.gguf | Q5_K_L | 90.91GB | true | Uses Q8_0 for embed and output weights. High quality, recommended. |

| Ling-3.0-flash-Q5_K_M.gguf | Q5_K_M | 90.66GB | true | High quality, recommended. |

| Ling-3.0-flash-Q5_K_S.gguf | Q5_K_S | 87.89GB | true | High quality, recommended. |

| Ling-3.0-flash-Q4_1.gguf | Q4_1 | 80.20GB | true | Legacy format, similar performance to Q4_K_S but with improved tokens/watt on Apple silicon. |

| Ling-3.0-flash-Q4_K_L.gguf | Q4_K_L | 78.10GB | true | Uses Q8_0 for embed and output weights. Good quality, recommended. |

| Ling-3.0-flash-Q4_K_M.gguf | Q4_K_M | 77.80GB | true | Good quality, default size for most use cases, recommended. |

| Ling-3.0-flash-Q4_K_S.gguf | Q4_K_S | 75.05GB | true | Slightly lower quality with more space savings, recommended. |

| Ling-3.0-flash-Q4_0.gguf | Q4_0 | 72.84GB | true | Legacy format, kept for compatibility with older tools. |

| Ling-3.0-flash-IQ4_NL.gguf | IQ4_NL | 72.56GB | true | Similar to IQ4_XS, but slightly larger. |

| Ling-3.0-flash-IQ4_XS.gguf | IQ4_XS | 68.74GB | true | Decent quality, smaller than Q4_K_S with similar performance, recommended. |

| Ling-3.0-flash-Q3_K_XL.gguf | Q3_K_XL | 61.78GB | true | Uses Q8_0 for embed and output weights. Lower quality but usable, good for low RAM availability. |

| Ling-3.0-flash-IQ3_M.gguf | IQ3_M | 61.52GB | true | Medium-low quality, new method with decent performance comparable to Q3_K_M. |

| Ling-3.0-flash-Q3_K_L.gguf | Q3_K_L | 61.43GB | true | Lower quality but usable, good for low RAM availability. |

| Ling-3.0-flash-Q3_K_M.gguf | Q3_K_M | 58.97GB | true | Low quality. |

| Ling-3.0-flash-IQ3_XS.gguf | IQ3_XS | 58.94GB | true | Lower quality, new method with decent performance, slightly better than Q3_K_S. |

| Ling-3.0-flash-Q3_K_S.gguf | Q3_K_S | 56.29GB | true | Low quality, not recommended. |

| Ling-3.0-flash-IQ3_XXS.gguf | IQ3_XXS | 54.12GB | true | Lower quality, new method with decent performance, comparable to Q3 quants. |

| Ling-3.0-flash-Q2_K_L.gguf | Q2_K_L | 46.29GB | false | Uses Q8_0 for embed and output weights. Very low quality but surprisingly usable. |

| Ling-3.0-flash-Q2_K.gguf | Q2_K | 45.90GB | false | Very low quality but surprisingly usable. |

| Ling-3.0-flash-IQ2_M.gguf | IQ2_M | 43.64GB | false | Relatively low quality, uses SOTA techniques to be surprisingly usable. |

| Ling-3.0-flash-IQ2_S.gguf | IQ2_S | 39.70GB | false | Low quality, uses SOTA techniques to be usable. |

| Ling-3.0-flash-IQ2_XS.gguf | IQ2_XS | 39.04GB | false | Low quality, uses SOTA techniques to be usable. |

| Ling-3.0-flash-IQ2_XXS.gguf | IQ2_XXS | 35.24GB | false | Very low quality, uses SOTA techniques to be usable. |

| Ling-3.0-flash-IQ1_M.gguf | IQ1_M | 30.56GB | false | Extremely low quality, not recommended. |

| Ling-3.0-flash-IQ1_S.gguf | IQ1_S | 27.57GB | false | Extremely low quality, not recommended. |

Download a specific file:

hf download bartowski/Ling-3.0-flash-GGUF --include "Ling-3.0-flash-Q4_K_M/*" --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/Ling-3.0-flash-GGUF --include "Ling-3.0-flash-Q4_K_M/*" --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/Ling-3.0-flash-GGUF --include "Ling-3.0-flash-Q8_0/*" --local-dir ./

You can either specify a new local-dir (Ling-3.0-flash-Q8_0) 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/Ling-3.0-flash-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 b10472 - 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

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 Q4_0 in the imatrix quants (except for the Q8_0 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: Ling-3.0-flash-calibration-v6.txt. The imatrix is available here: Ling-3.0-flash-imatrix.gguf.

<details>

<summary>Calibration render details</summary>

{
  "generator": "auto_quant_v2 calibration renderer",
  "recipe": "calibration-v6",
  "model": "Ling-3.0-flash",
  "encoder": "chat_template",
  "chunk_size": 512,
  "prose_chunks": 220,
  "tool_chunks": 345,
  "total_chunks": 565,
  "tool_chunk_fraction": 0.611,
  "n_conversations": 137,
  "extension_convs_used": 0,
  "conversation_token_lengths": [
    523,
    1594,
    1193,
    1476,
    1046,
    1300,
    3127,
    754,
    1163,
    1353,
    1019,
    2059,
    836,
    1200,
    2755,
    1189,
    1099,
    948,
    694,
    677,
    1326,
    990,
    1308,
    1167,
    1839,
    1463,
    1601,
    844,
    1376,
    1604,
    1472,
    1161,
    1211,
    1003,
    1019,
    1650,
    1619,
    1147,
    433,
    1912,
    1392,
    1048,
    1355,
    1973,
    2023,
    1230,
    1569,
    824,
    2903,
    1063,
    2811,
    723,
    955,
    915,
    924,
    655,
    2396,
    840,
    1100,
    1045,
    1166,
    1133,
    868,
    1151,
    1114,
    1530,
    873,
    1483,
    2099,
    803,
    333,
    1071,
    3285,
    2856,
    671,
    865,
    974,
    1022,
    1244,
    1052,
    1074,
    753,
    1152,
    983,
    1244,
    1468,
    1321,
    2041,
    795,
    608,
    2714,
    658,
    1345,
    1626,
    1936,
    1168,
    581,
    1336,
    1136,
    1653,
    1759,
    1625,
    782,
    961,
    976,
    2730,
    697,
    679,
    709,
    1354,
    1011,
    1544,
    731,
    361,
    327,
    2569,
    947,
    1085,
    1815,
    1970,
    2651,
    2644,
    759,
    931,
    797,
    884,
    1190,
    944,
    809,
    1266,
    793,
    668,
    1711,
    965,
    880,
    1240,
    1409
  ],
  "warnings": []
}

</details>

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L 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 Q4_0_4_4/4_8/8_8 files; those are long gone. Online repacking now covers Q4_0, 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 'QX_K_X', like Q5_K_M.

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 IQX_X, like IQ3_M. 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

Run bartowski/Ling-3.0-flash-GGUF with guIDE

Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.

Download guIDE → · Browse 524k+ models · Compare models

Source: Hugging Face · Compare models