Skip to content
localllm
日本語

inclusionAI

Ling-3.0-flash

Parameters
124.4B
Max context length
131,072
Architecture
bailingmoe3
License
Commercial OK
Base model
inclusionAI/Ling-3.0-flash
License
mit
llama.cpp
b10470+

About Ling-3.0-flash

Ling-3.0-flash is a GGUF model based on inclusionAI/Ling-3.0-flash on Hugging Face. GGUF files are published by 2 sources (AtomicChat/Ling-3.0-flash-GGUF, bloomer010/Ling-3.0-flash-GGUF). It has about 124.4B parameters, the bailingmoe3 architecture, a maximum context length of 131,072 tokens (values written in the GGUF).

There are 31 quantizations, with file sizes from 19.5 GB (Q1_0) to 255 GB (BF16) (actual sizes on Hugging Face).

Running Q4_K_M (75.3 GB) with an 8,192-token context needs an estimated 76.3 GB of memory, and an estimated 76.4 GB at 32,768 tokens. For 8,192 tokens, a GPU with 96 GB or more of VRAM, or a Mac with 96 GB or more of unified memory, is a rough guide (macOS caps how much memory the GPU can use, so leave some headroom). The smallest, Q1_0 (19.5 GB), needs an estimated 20.5 GB at 8,192 tokens.

The original model is licensed under MIT License, which allows commercial use (keep the license text and copyright notice when redistributing). Check the original license for the terms of use.

File size and estimated memory per quantization

  • Up to 8 GB
  • Up to 16 GB
  • Up to 24 GB
  • Up to 48 GB
  • Up to 128 GB
  • More

File sizes are the actual values on Hugging Face. Memory is estimated for each context length (KV cache f16). The formula is in How memory is calculated.

All memory figures are Estimated (KV cache f16). It should run on a setup with at least this much VRAM (or Mac unified memory). Click a value to see the devices and models that fit that amount (Find by GPU / Mac).

2 sources publish GGUF files of this model. Quantization types and file sizes differ slightly between sources, so there is a table for each source below.

About the "quality guide" (how far a quantization drifts from the original)

For each quantization type, how far the output drifts from the original model, grouped into six levels from values llama.cpp measured on Llama-3-8B.

The value is mean KLD (the difference between the original and quantized model in next-token probabilities); smaller means closer to the original. Quantizations made with an imatrix (importance data) come out slightly smaller. Levels use the value without an imatrix (or with one, for types that only have that).

These values were not measured on this model. Larger models and MoE models tend to drift less with the same type. Distributor-specific mixed types (such as UD-Q4_K_XL) cannot be matched by type name alone, so they show "-".

Quantization Quality guide KLD (no imatrix) KLD (imatrix) Size (Llama-3-8B)
Q8_0 Nearly identical 0.0014 - 7.96 GiB
Q6_K Very small 0.0055 - 6.14 GiB
Q5_K_M Very small 0.0108 - 5.33 GiB
Q5_K_S Small 0.0166 - 5.21 GiB
Q5_1 Small 0.0180 - 5.65 GiB
Q5_0 Small 0.0222 - 5.21 GiB
Q4_K_M Small 0.0313 0.0282 4.58 GiB
Q4_K_S Medium 0.0431 0.0320 4.37 GiB
IQ4_NL Small - 0.0357 4.35 GiB
IQ4_XS Small - 0.0363 4.14 GiB
Q4_1 Medium 0.0717 - 4.78 GiB
Q4_0 Medium 0.0719 - 4.34 GiB
Q3_K_L Medium 0.0872 0.0731 4.03 GiB
Q3_K_M Medium 0.1019 0.0844 3.74 GiB
IQ3_M Medium - 0.1025 3.53 GiB
IQ3_S Medium - 0.1113 3.42 GiB
IQ3_XS Large - 0.1387 3.28 GiB
IQ3_XXS Large - 0.1836 3.05 GiB
Q3_K_S Large 0.2282 0.1998 3.41 GiB
IQ2_M Large - 0.3260 2.74 GiB
Q2_K Very large 0.4451 0.3322 2.96 GiB
Q2_K_S Very large - 0.4034 2.96 GiB
IQ2_S Very large - 0.4392 2.56 GiB
IQ2_XS Very large - 0.5463 2.43 GiB
IQ2_XXS Very large - 0.8120 2.24 GiB
IQ1_M Very large - 1.3931 2.01 GiB
IQ1_S Very large - 2.2546 1.88 GiB

Source: llama.cpp (MIT), measured on Llama-3-8B against FP16, Wikitext-2, revision f364eb6f. Original table · Reading and choosing quantizations

GGUF source

AtomicChat/Ling-3.0-flash-GGUF

Converted and published by a third party About this label

Last checked: Nov 11, 2026

Architecture
bailingmoe3
llama.cpp b10470 and later
Parameters
124.4B
Max context length
131,072
License
mit Commercial OK
File size and memory per quantization for Ling-3.0-flash (AtomicChat/Ling-3.0-flash-GGUF)
Quantization File size Measured Quality guide 4,096 8,192 16,384 32,768 65,536 131,072 max
IQ1_S 32.4 GB Very large 33.4 GB 33.4 GB 33.4 GB 33.5 GB 33.8 GB 34.2 GB
IQ1_M 36.5 GB Very large 37.4 GB 37.5 GB 37.5 GB 37.6 GB 37.9 GB 38.3 GB
IQ2_XXS 39.2 GB Very large 40.2 GB 40.2 GB 40.3 GB 40.4 GB 40.6 GB 41.1 GB
IQ2_XS 44.7 GB Very large 45.7 GB 45.7 GB 45.8 GB 45.9 GB 46.1 GB 46.6 GB
IQ2_S (2 parts) 46.9 GB Very large 47.9 GB 47.9 GB 48.0 GB 48.1 GB 48.3 GB 48.8 GB
IQ2_M (2 parts) 49.1 GB Large 50.1 GB 50.1 GB 50.1 GB 50.3 GB 50.5 GB 50.9 GB
IQ3_XXS (2 parts) 57.1 GB Large 58.0 GB 58.1 GB 58.1 GB 58.2 GB 58.5 GB 58.9 GB
IQ3_S (2 parts) 57.8 GB Medium 58.8 GB 58.8 GB 58.9 GB 59.0 GB 59.2 GB 59.7 GB
IQ3_M (2 parts) 62.2 GB Medium 63.2 GB 63.2 GB 63.2 GB 63.4 GB 63.6 GB 64.0 GB
IQ4_XS (2 parts) IQ4_XS_STOCK 66.4 GB Small 67.4 GB 67.4 GB 67.5 GB 67.6 GB 67.8 GB 68.3 GB
IQ4_XS (2 parts) IQ4_XS_FLAT 68.6 GB Small 69.5 GB 69.6 GB 69.6 GB 69.7 GB 70.0 GB 70.4 GB
IQ4_XXS (2 parts) 69.3 GB - 70.3 GB 70.3 GB 70.4 GB 70.5 GB 70.7 GB 71.2 GB
NVFP4 (2 parts) NVFP4_STOCK 72.3 GB - 73.3 GB 73.3 GB 73.4 GB 73.5 GB 73.7 GB 74.2 GB
NVFP4 (2 parts) AD-NVFP4_AI2 72.3 GB - 73.3 GB 73.3 GB 73.4 GB 73.5 GB 73.7 GB 74.2 GB
NVFP4 (2 parts) AD-NVFP4_V5 72.3 GB - 73.3 GB 73.3 GB 73.4 GB 73.5 GB 73.7 GB 74.2 GB
Q4_K (2 parts) 72.3 GB - 73.3 GB 73.3 GB 73.4 GB 73.5 GB 73.7 GB 74.2 GB
Q4_K_S (2 parts) 74.2 GB Medium 75.2 GB 75.2 GB 75.3 GB 75.4 GB 75.6 GB 76.1 GB
IQ4_XS (2 parts) AD-IQ4_XS 74.9 GB Small 75.8 GB 75.8 GB 75.9 GB 76.0 GB 76.2 GB 76.7 GB
Q4_K_M (2 parts) Q4_K_M_STOCK 75.3 GB Small 76.2 GB 76.3 GB 76.3 GB 76.4 GB 76.7 GB 77.1 GB
IQ4_NL (2 parts) 79.3 GB Small 80.3 GB 80.3 GB 80.3 GB 80.5 GB 80.7 GB 81.1 GB
Q4_K_M (2 parts) AD-Q4_K_M 79.3 GB Small 80.3 GB 80.3 GB 80.3 GB 80.5 GB 80.7 GB 81.1 GB
Q4_K_L (2 parts) 84.0 GB - 84.9 GB 85.0 GB 85.0 GB 85.1 GB 85.4 GB 85.8 GB
Q5_K_S (2 parts) 87.4 GB Small 88.4 GB 88.4 GB 88.5 GB 88.6 GB 88.8 GB 89.3 GB
Q5_K_M (2 parts) Q5_K_M_STOCK 88.3 GB Very small 89.3 GB 89.3 GB 89.4 GB 89.5 GB 89.7 GB 90.2 GB
Q5_K_M (2 parts) AD-Q5_K_M 89.4 GB Very small 90.4 GB 90.4 GB 90.5 GB 90.6 GB 90.8 GB 91.3 GB
Q5_K_L (3 parts) 95.3 GB - 96.3 GB 96.3 GB 96.4 GB 96.5 GB 96.7 GB 97.2 GB
Q6_K (3 parts) Q6_K_STOCK 102 GB Very small 103 GB 103 GB 103 GB 103 GB 104 GB 104 GB
Q6_K (3 parts) AD-Q6_K 108 GB Very small 108 GB 109 GB 109 GB 109 GB 109 GB 109 GB
Q8_0 (3 parts) 133 GB Nearly identical 134 GB 134 GB 134 GB 134 GB 135 GB 135 GB
BF16 (6 parts) 249 GB Original precision 250 GB 250 GB 250 GB 250 GB 250 GB 251 GB 
Ollama  ollama run hf.co/AtomicChat/Ling-3.0-flash-GGUF:Q4_K_M
llama.cpp  llama-server -hf AtomicChat/Ling-3.0-flash-GGUF:Q4_K_M

GGUF source

bloomer010/Ling-3.0-flash-GGUF

Converted and published by a third party About this label

Last checked: Nov 11, 2026

Architecture
bailingmoe3
llama.cpp b10470 and later
Parameters
127.5B
Max context length
262,144
License
mit Commercial OK
File size and memory per quantization for Ling-3.0-flash (bloomer010/Ling-3.0-flash-GGUF)
Quantization File size Measured Quality guide 4,096 8,192 16,384 32,768 65,536 131,072 262,144 max
Q1_0 19.5 GB - 20.4 GB 20.5 GB 20.5 GB 20.7 GB 20.9 GB 21.5 GB 22.5 GB
IQ1_S 26.7 GB Very large 27.7 GB 27.7 GB 27.8 GB 27.9 GB 28.2 GB 28.7 GB 29.7 GB
IQ1_M 29.5 GB Very large 30.5 GB 30.5 GB 30.6 GB 30.7 GB 31.0 GB 31.5 GB 32.6 GB
IQ2_M 42.1 GB Large 43.1 GB 43.1 GB 43.2 GB 43.3 GB 43.6 GB 44.1 GB 45.1 GB
UD-Q2_K_XL 43.4 GB - 44.4 GB 44.4 GB 44.5 GB 44.6 GB 44.9 GB 45.4 GB 46.5 GB
IQ3_XXS 51.2 GB Large 52.2 GB 52.2 GB 52.3 GB 52.4 GB 52.7 GB 53.2 GB 54.3 GB
Q3_K_M 62.6 GB Medium 63.5 GB 63.6 GB 63.6 GB 63.8 GB 64.0 GB 64.6 GB 65.6 GB
UD-Q3_K_XL 62.6 GB - 63.6 GB 63.6 GB 63.7 GB 63.8 GB 64.1 GB 64.6 GB 65.7 GB
MXFP4_MOE 69.8 GB - 70.8 GB 70.9 GB 70.9 GB 71.0 GB 71.3 GB 71.8 GB 72.9 GB
Q4_K_S 73.9 GB Medium 74.9 GB 74.9 GB 75.0 GB 75.1 GB 75.4 GB 75.9 GB 77.0 GB
Q4_K_M 78.3 GB Small 79.3 GB 79.3 GB 79.4 GB 79.5 GB 79.8 GB 80.3 GB 81.3 GB
UD-Q4_K_XL 84.7 GB - 85.7 GB 85.7 GB 85.8 GB 85.9 GB 86.2 GB 86.7 GB 87.8 GB
Q5_K_S 89.0 GB Small 90.0 GB 90.0 GB 90.1 GB 90.2 GB 90.5 GB 91.0 GB 92.1 GB
Q5_K_M 91.5 GB Very small 92.5 GB 92.5 GB 92.6 GB 92.7 GB 92.9 GB 93.5 GB 94.5 GB
Q6_K 106 GB Very small 106 GB 107 GB 107 GB 107 GB 107 GB 108 GB 109 GB
UD-Q6_K_XL 116 GB - 117 GB 117 GB 117 GB 117 GB 117 GB 118 GB 119 GB
Q8_0 136 GB Nearly identical 137 GB 137 GB 137 GB 137 GB 137 GB 138 GB 139 GB
UD-Q8_K_XL 177 GB - 178 GB 178 GB 178 GB 178 GB 179 GB 179 GB 180 GB
BF16 255 GB Original precision 256 GB 256 GB 256 GB 256 GB 257 GB 257 GB 258 GB 
Ollama  ollama run hf.co/bloomer010/Ling-3.0-flash-GGUF:Q4_K_M
llama.cpp  llama-server -hf bloomer010/Ling-3.0-flash-GGUF:Q4_K_M

License and commercial use

MIT License Commercial OK

Allows commercial use, modification and redistribution. When redistributing, keep the copyright notice and license text.

Based on the license on the original model's Hugging Face page. Original license text (checked Oct 18, 2026)

The summary states facts this site read from the original text and is not legal advice. Always check the original license and its latest terms before use. All licenses

How to run

Mainline llama.cpp can load this model's format (bailingmoe3) from b10470 (Aug 17, 2026) onward. Ollama and LM Studio bundle llama.cpp internally, so an older app fails to load it with errors such as "unknown model architecture". If that happens, update the app.

Example commands using Q4_K_M. To use another quantization, replace the name with one from the table above. Commands for files from other sources are in each source's section. How to run local LLMs

Ollama

Run it by pointing directly at the GGUF on Hugging Face.

ollama run hf.co/AtomicChat/Ling-3.0-flash-GGUF:Q4_K_M

llama.cpp

Pass the repository and quantization name to llama-server (or llama-cli) with -hf.

llama-server -hf AtomicChat/Ling-3.0-flash-GGUF:Q4_K_M

LM Studio

Paste AtomicChat/Ling-3.0-flash-GGUF into the search box, download it, and choose the quantization from the file list.

Frequently asked questions

How much memory do I need to run Ling-3.0-flash (Q4_K_M)?
The file size is 75.3 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 76.3 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Can Ling-3.0-flash be used commercially?
The original model's license is MIT License (Commercial use allowed).Allows commercial use, modification and redistribution. When redistributing, keep the copyright notice and license text. Check the original license text before deciding.
Does Ling-3.0-flash run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?
8 GB: no quantization fits; 12 GB: no quantization fits; 16 GB: no quantization fits; 24 GB: no quantization fits (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run Ling-3.0-flash with Ollama?
Run "ollama run hf.co/AtomicChat/Ling-3.0-flash-GGUF:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.