Skip to content
localllm
日本語

knifeayumu

LLM_Collection

Parameters
32.3B
Max context length
131,072
Architecture
command-r

About LLM_Collection

LLM_Collection is a model published in GGUF format. GGUF files are published by knifeayumu/LLM_Collection. It has about 32.3B parameters, the command-r architecture, a maximum context length of 131,072 tokens (values written in the GGUF).

There are 8 quantizations, with file sizes from 3.29 GB (Q8_0) to 73.5 GB (Q4_K_M) (actual sizes on Hugging Face).

Running Q4_K_M (41.4 GB) with an 8,192-token context needs an estimated 43.9 GB of memory, and an estimated 49.5 GB at 32,768 tokens. For 8,192 tokens, a GPU with 48 GB or more of VRAM, or a Mac with 48 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, Q8_0 (3.29 GB), needs an estimated 5.77 GB at 8,192 tokens.

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).

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

knifeayumu/LLM_Collection

Published by the original developer About this label

Last checked: Oct 19, 2026

Architecture
command-r
Supported before llama.cpp b4409
Parameters
32.3B
Max context length
131,072
File size and memory per quantization for LLM_Collection (knifeayumu/LLM_Collection)
Quantization File size Measured Quality guide 4,096 8,192 16,384 32,768 65,536 131,072 max
Q8_0 Kwen-3B-v1d-Q8_0 3.29 GB Nearly identical 4.83 GB 5.77 GB 7.65 GB 11.4 GB 18.9 GB 34.0 GB
F16 5.24 GB Original precision 6.78 GB 7.72 GB 9.60 GB 13.4 GB 20.9 GB 35.9 GB
Q6_K Toppy-M-7B.q6_k 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K mistral-7b-instruct-v0.1.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K WestLake-7B-v2-Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K openhermes-2.5-neural-chat-7b-v3-1-7b.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K Fett-uccine-7B-Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K ShoriRP.v077.q6_k 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K mistral-7b-instruct-v0.2.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K Nous-Hermes-2-Mistral-7B-DPO.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K neuralhermes-2.5-mistral-7b.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K openchat-3.5-0106.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K Starling-LM-7B-beta-Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K mistral-7b-v0.1-layla-v4-chatml-Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K mistral-7b-v0.2-layla-v4-Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K openhermes-2.5-neural-chat-7b-v3-2-7b.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K kunoichi-7b.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q6_K silicon-maid-7b.Q6_K 5.94 GB Very small 7.49 GB 8.43 GB 10.3 GB 14.1 GB 21.6 GB 36.6 GB
Q8_0 kunoichi-dpo-v2-7b.Q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 neural-chat-7b-v3-1-16k-q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 neural-chat-7b-v3-16k-q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 LemonadeRP-4.5.3.Q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 LemonWizardv3.Q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 WizardLM-2-7B.Q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 Erosumika-7B.q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 LemonKunoichiWizardV3.Q8_0 7.70 GB Nearly identical 9.24 GB 10.2 GB 12.1 GB 15.8 GB 23.3 GB 38.4 GB
Q8_0 Ministrations-8B-v1a-Q8_0 8.53 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.6 GB 24.2 GB 39.2 GB
Q8_0 Llama-3-Lumimaid-8B-v0.1-OAS.q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 llama3-8b-cunnygpt-16bit-q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 Llama-3SOME-8B-v2-Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 L3-8B-Lunaris-v1-Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 L3-8B-Stheno-v3.2-Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 L3-8B-Sunfall-v0.4-Stheno-v3.2-Q8_0-imat 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 L3-8B-sunfall-v0.4-stheno-v3.2.Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 Llama-3.1-8B-ArliAI-RPMax-v1.2-q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 Llama-3-Soliloquy-8B-v2.Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 L3.1-8B-sunfall-stheno-v0.6.1-q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 Meta-Llama-3.1-8B-Instruct-Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 L3.1-8B-sunfall-v0.6.1-dpo.Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q8_0 Llama-3some-8B-v1-rc1-Q8_0 8.54 GB Nearly identical 10.1 GB 11.0 GB 12.9 GB 16.7 GB 24.2 GB 39.2 GB
Q6_K Fimbulvetr-10.7B-v1.q6_K 8.81 GB Very small 10.3 GB 11.3 GB 13.2 GB 16.9 GB 24.4 GB 39.5 GB
Q5_K_M LLaMA2-13B-Psyfighter2.Q5_K_M 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q5_K_M LLaMA2-13B-Tiefighter.Q5_K_M 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q5_K_M utopia-13b.Q5_K_M 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q5_K_M NeteLegacy-13B.q5_k_m 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q5_K_M Xwin-MLewd-13B-V0.2.q5_k_m 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q5_K_M Noromaid-13b-v0.1.1.q5_k_m 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q5_K_M Noromaid-13b-v0.2.q5_k_m 9.23 GB Very small 10.8 GB 11.7 GB 13.6 GB 17.4 GB 24.9 GB 39.9 GB
Q8_0 Gemmasutra-9B-v1c-Q8_0 9.83 GB Nearly identical 11.4 GB 12.3 GB 14.2 GB 17.9 GB 25.5 GB 40.5 GB
Q8_0 gemma-2-9b-it-Q8_0 9.83 GB Nearly identical 11.4 GB 12.3 GB 14.2 GB 17.9 GB 25.5 GB 40.5 GB
Q8_0 Fimbulvetr-11B-v2-Test-14.q8_0 11.4 GB Nearly identical 12.9 GB 13.9 GB 15.8 GB 19.5 GB 27.0 GB 42.1 GB
Q8_0 Moistral-11B-v3-Q8_0 11.4 GB Nearly identical 12.9 GB 13.9 GB 15.8 GB 19.5 GB 27.0 GB 42.1 GB
Q8_0 Fimbulvetr-11B-v2.1-16K.Q8_0 11.4 GB Nearly identical 12.9 GB 13.9 GB 15.8 GB 19.5 GB 27.0 GB 42.1 GB
Q8_0 Sunfall-NemoMix-Unleashed-12B-v0.6.1-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Rocinante-12B-v1.1-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Rocinante-12B-v2j-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 NemoRemix-12B-q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 NemoReRemix-12B-q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 NemoMix-Unleashed-12B-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 magnum-12b-v2-q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Chronos-Gold-12B-1.0-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 nemo-sunfall-v0.6.1-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Mistral-Nemo-Instruct-2407-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Rocinante-12B-v1-nemo-sunfall-v0.6.1-SLERP-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Rocinante-12B-v2i-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Mistral-Nemo-12B-ArliAI-RPMax-v1.2-q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 Rocinante-12B-v1-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q8_0 magnum-v4-12b-Q8_0 13.0 GB Nearly identical 14.6 GB 15.5 GB 17.4 GB 21.1 GB 28.7 GB 43.7 GB
Q5_K_M Emerhyst-20B.q5_k_m 14.2 GB Very small 15.7 GB 16.6 GB 18.5 GB 22.3 GB 29.8 GB 44.8 GB
Q5_K_M psymedrp-v1-20b.Q5_K_M 14.2 GB Very small 15.7 GB 16.6 GB 18.5 GB 22.3 GB 29.8 GB 44.8 GB
Q5_K_M u-amethyst-20b.Q5_K_M 14.2 GB Very small 15.7 GB 16.6 GB 18.5 GB 22.3 GB 29.8 GB 44.8 GB
Q5_K_M psyonic-cetacean-20b.Q5_K_M 14.2 GB Very small 15.7 GB 16.6 GB 18.5 GB 22.3 GB 29.8 GB 44.8 GB
Q5_K_M Noromaid-20b-v0.1.1.q5_k_m 14.2 GB Very small 15.7 GB 16.6 GB 18.5 GB 22.3 GB 29.8 GB 44.8 GB
Q5_K_M rose-20b.Q5_K_M 14.2 GB Very small 15.7 GB 16.6 GB 18.5 GB 22.3 GB 29.8 GB 44.8 GB
Q5_K_M ChaoticSoliloquy-v1.5-4x8B-Q5_K_M 17.7 GB Very small 19.3 GB 20.2 GB 22.1 GB 25.9 GB 33.4 GB 48.4 GB
Q8_0 Theia-21B-v1-Q8_0 21.7 GB Nearly identical 23.3 GB 24.2 GB 26.1 GB 29.8 GB 37.4 GB 52.4 GB
Q8_0 Theia-21B-v2b-Q8_0 21.7 GB Nearly identical 23.3 GB 24.2 GB 26.1 GB 29.8 GB 37.4 GB 52.4 GB
Q8_0 reka-flash-3-Q8_0 22.2 GB Nearly identical 23.8 GB 24.7 GB 26.6 GB 30.3 GB 37.9 GB 52.9 GB
Q8_0 Codestral-22B-v0.1-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-22B-v1a-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-22B-v1e-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-22B-v1f-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Mistral-Small-Instruct-2409-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 MS-Sunfall-v0.7.0-q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-22B-v2k-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Mistral-Small-22B-ArliAI-RPMax-v1.1-q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-22B-v2f-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-22B-v2m-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 magnum-v4-22b-Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-v1.2-v1.3-Magnum-v4-22B.Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-v1.2-v1.3-22B.Q8_0 23.6 GB Nearly identical 25.2 GB 26.1 GB 28.0 GB 31.8 GB 39.3 GB 54.3 GB
Q8_0 Cydonia-24B-v2g-Q8_0 25.1 GB Nearly identical 26.6 GB 27.5 GB 29.4 GB 33.2 GB 40.7 GB 55.7 GB
Q8_0 Mistral-Small-24B-Instruct-2501-Q8_0 25.1 GB Nearly identical 26.6 GB 27.5 GB 29.4 GB 33.2 GB 40.7 GB 55.7 GB
Q8_0 Mistral-Small-3.1-24B-Instruct-2503-Q8_0 25.1 GB Nearly identical 26.6 GB 27.5 GB 29.4 GB 33.2 GB 40.7 GB 55.7 GB
Q8_0 Cydonia-24B-v2l-Q8_0 25.1 GB Nearly identical 26.6 GB 27.5 GB 29.4 GB 33.2 GB 40.7 GB 55.7 GB
Q8_0 TheDrummer_Cydonia-24B-v4-Q8_0 25.1 GB Nearly identical 26.6 GB 27.5 GB 29.4 GB 33.2 GB 40.7 GB 55.7 GB
Q8_0 Fallen-Gemma3-27B-v1c-Q8_0 28.7 GB Nearly identical 30.3 GB 31.2 GB 33.1 GB 36.8 GB 44.3 GB 59.4 GB
Q8_0 gemma-3-27b-it.Q8_0 28.7 GB Nearly identical 30.3 GB 31.2 GB 33.1 GB 36.8 GB 44.3 GB 59.4 GB
Q8_0 gemma-2-27b-it-Q8_0 28.9 GB Nearly identical 30.5 GB 31.4 GB 33.3 GB 37.1 GB 44.6 GB 59.6 GB
Q8_0 Gemmasutra-Pro-27B-v1i-Q8_0 28.9 GB Nearly identical 30.5 GB 31.4 GB 33.3 GB 37.1 GB 44.6 GB 59.6 GB
Q8_0 magnum-v3-27b-kto-Q8_0 28.9 GB Nearly identical 30.5 GB 31.4 GB 33.3 GB 37.1 GB 44.6 GB 59.6 GB
Q8_0 TheDrummer_Big-Tiger-Gemma-27B-v3-Q8_0 30.2 GB Nearly identical 31.7 GB 32.7 GB 34.6 GB 38.3 GB 45.8 GB 60.9 GB
Q5_K_M flatdolphinmaid-8x7b.Q5_K_M 32.2 GB Very small 33.8 GB 34.7 GB 36.6 GB 40.4 GB 47.9 GB 62.9 GB
Q5_K_M Noromaid-v0.4-Mixtral-Instruct-8x7b.q5_k_m 32.2 GB Very small 33.8 GB 34.7 GB 36.6 GB 40.4 GB 47.9 GB 62.9 GB
Q5_K_M BagelMIsteryTour-v2-8x7B.Q5_K_M.imx 33.2 GB Very small 34.8 GB 35.7 GB 37.6 GB 41.3 GB 48.9 GB 63.9 GB
Q5_K_M maid-yuzu-v8-alter-Q5_K_M 33.2 GB Very small 34.8 GB 35.7 GB 37.6 GB 41.3 GB 48.9 GB 63.9 GB
Q5_K_M Mixtral-8x7B-Instruct-v0.1-Q5_K_M 33.2 GB Very small 34.8 GB 35.7 GB 37.6 GB 41.3 GB 48.9 GB 63.9 GB
Q5_K_M Nous-Hermes-2-Mixtral-8x7B-DPO.Q5_K_M 33.2 GB Very small 34.8 GB 35.7 GB 37.6 GB 41.3 GB 48.9 GB 63.9 GB
Q8_0 OLMo-2-0325-32B-Instruct-Q8_0 34.3 GB Nearly identical 35.8 GB 36.7 GB 38.6 GB 42.4 GB 49.9 GB 64.9 GB
Q8_0 Star-Command-R-32B-v1-Q8_0 34.3 GB Nearly identical 35.9 GB 36.8 GB 38.7 GB 42.4 GB 50.0 GB 65.0 GB
Q8_0 Star-Command-32B-v2b-Q8_0 34.3 GB Nearly identical 35.9 GB 36.8 GB 38.7 GB 42.4 GB 50.0 GB 65.0 GB
Q8_0 c4ai-command-r-08-2024-Q8_0 34.3 GB Nearly identical 35.9 GB 36.8 GB 38.7 GB 42.4 GB 50.0 GB 65.0 GB
Q8_0 GLM-4-32B-0414-Q8_0 34.6 GB Nearly identical 36.2 GB 37.1 GB 39.0 GB 42.7 GB 50.3 GB 65.3 GB
Q8_0 Qwen3-32B-Q8_0 34.8 GB Nearly identical 36.4 GB 37.3 GB 39.2 GB 42.9 GB 50.5 GB 65.5 GB
Q8_0 Qwen2.5-32B-Instruct-Q8_0 34.8 GB Nearly identical 36.4 GB 37.3 GB 39.2 GB 42.9 GB 50.5 GB 65.5 GB
Q8_0 QwQ-32B-Q8_0 34.8 GB Nearly identical 36.4 GB 37.3 GB 39.2 GB 42.9 GB 50.5 GB 65.5 GB
Q8_0 Fallen-QwQ-32B-v1d-Q8_0 34.8 GB Nearly identical 36.4 GB 37.3 GB 39.2 GB 42.9 GB 50.5 GB 65.5 GB
Q8_0 Coomand-R-35B-v1-Q8_0 37.2 GB Nearly identical 38.7 GB 39.7 GB 41.5 GB 45.3 GB 52.8 GB 67.8 GB
Q8_0 c4ai-command-r-v01.Q8_0 37.2 GB Nearly identical 38.7 GB 39.7 GB 41.5 GB 45.3 GB 52.8 GB 67.8 GB
Q6_K mixtral-8x7b-instruct-v0.1-limarp-zloss.Q6_K 38.4 GB Very small 39.9 GB 40.9 GB 42.7 GB 46.5 GB 54.0 GB 69.0 GB
Q6_K NoromaidxOpenGPT4-2.Q6_K 38.5 GB Very small 40.1 GB 41.0 GB 42.9 GB 46.7 GB 54.2 GB 69.2 GB
Q6_K TheDrummer_Valkyrie-49B-v1a-Q6_K 40.9 GB Very small 42.5 GB 43.4 GB 45.3 GB 49.0 GB 56.6 GB 71.6 GB
Q6_K Valkyrie-49B-v1a-Q6_K 40.9 GB Very small 42.5 GB 43.4 GB 45.3 GB 49.0 GB 56.6 GB 71.6 GB
Q4_K_M ggml-senku-70b-q4_k_m 41.4 GB Small 43.0 GB 43.9 GB 45.8 GB 49.5 GB 57.1 GB 72.1 GB
Q4_K_M Miqu-70B-DPO.q4_k_m 41.4 GB Small 43.0 GB 43.9 GB 45.8 GB 49.5 GB 57.1 GB 72.1 GB
Q4_K_M OnlyChat-Miqu-v1.q4_k_m 41.4 GB Small 43.0 GB 43.9 GB 45.8 GB 49.5 GB 57.1 GB 72.1 GB
Q4_K_M limarp-miqu-1-70b-Q4_K_M 41.4 GB Small 43.0 GB 43.9 GB 45.8 GB 49.5 GB 57.1 GB 72.1 GB
Q4_K_M MiquMaid-v3-70B.Q4_K_M 41.4 GB Small 43.0 GB 43.9 GB 45.8 GB 49.5 GB 57.1 GB 72.1 GB
Q4_K_M Moist-Miqu-70B-v1-RC1-60p-Q4_K_M 41.4 GB Small 43.0 GB 43.9 GB 45.8 GB 49.5 GB 57.1 GB 72.1 GB
Q4_K_M Midnight-Miqu-70B-v1.5.Q4_K_M 41.7 GB Small 43.3 GB 44.2 GB 46.1 GB 49.9 GB 57.4 GB 72.4 GB
Q4_K_M Llama-3-Lumimaid-70B-v0.1-OAS.q4_k_m 42.5 GB Small 44.1 GB 45.0 GB 46.9 GB 50.6 GB 58.2 GB 73.2 GB
Q4_K_M Athene-70B.Q4_K_M 42.5 GB Small 44.1 GB 45.0 GB 46.9 GB 50.6 GB 58.2 GB 73.2 GB
Q4_K_M L3-70B-Euryale-v2.1.i1-Q4_K_M 42.5 GB Small 44.1 GB 45.0 GB 46.9 GB 50.6 GB 58.2 GB 73.2 GB
Q4_K_M Llama-3-Soliloquy-Max-70B-v1.Q4_K_M 42.5 GB Small 44.1 GB 45.0 GB 46.9 GB 50.6 GB 58.2 GB 73.2 GB
IQ3_M 47.1 GB Medium 48.7 GB 49.6 GB 51.5 GB 55.2 GB 62.8 GB 77.8 GB
Q5_K_M miqu-1-70b.q5_K_M 48.8 GB Very small 50.3 GB 51.2 GB 53.1 GB 56.9 GB 64.4 GB 79.4 GB
Q8_0 (2 parts) Llama-3.3-70B_49B 53.0 GB Nearly identical 54.5 GB 55.5 GB 57.4 GB 61.1 GB 68.6 GB 83.7 GB
Q5_K_M (2 parts) Qwen2.5-72B/magnum-v4-72b-Q5_K_M 54.4 GB Very small 56.0 GB 56.9 GB 58.8 GB 62.6 GB 70.1 GB 85.1 GB
Q5_K_M (2 parts) Qwen2.5-72B 54.4 GB Very small 56.0 GB 56.9 GB 58.8 GB 62.6 GB 70.1 GB 85.1 GB
IQ4_XS (2 parts) Llama-3.3-70B_105B 56.1 GB Small 57.7 GB 58.6 GB 60.5 GB 64.3 GB 71.8 GB 86.8 GB
Q6_K (2 parts) Miqu-1-70B/Nimbus-Miqu-v0.1-70B.Q6_K 56.6 GB Very small 58.1 GB 59.1 GB 60.9 GB 64.7 GB 72.2 GB 87.3 GB
Q6_K (2 parts) Miqu-1-70B/sunfall-midnight-miqu-v0.2-v1.5-70B.Q6_K.gguf 56.6 GB Very small 58.1 GB 59.1 GB 60.9 GB 64.7 GB 72.2 GB 87.3 GB
Q6_K (2 parts) Miqu-1-70B 56.6 GB Very small 58.1 GB 59.1 GB 60.9 GB 64.7 GB 72.2 GB 87.3 GB
Q6_K (2 parts) Llama-3-70B 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-70B/Llama-3.1-70B-ArliAI-RPMax-v1.1-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-70B/Llama-3.1-70B-ArliAI-RPMax-v1.2-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-70B 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/Negative_LLAMA_70B-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/Legion-V2.1-LLaMa-70B-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/L33-Tiger-R1-70B-v1b-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-Nemotron-70B/Nautilus-70B-v1a-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-70B/Meta-Llama-3.1-70B-Instruct-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/TheDrummer_Anubis-70B-v1.1-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-Nemotron-70B/Llama-3.1-Nemotron-70B-Instruct-HF-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/Llama-3.3-70B-Instruct-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/L3.3-70B-Euryale-v2.3.Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/Negative-Anubis-70B-v1-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/The-Omega-Directive-L-70B-v1.0.Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-Nemotron-70B/Llama-3.1-Nemotron-lorablated-70B.Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.1-Nemotron-70B 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/Electranova-70B-v1.0.Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/L3.3-MS-Nevoria-70b-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
Q6_K (2 parts) Llama-3.3-70B/L3.3-70B-Magnum-v4-SE-Q6_K 57.9 GB Very small 59.4 GB 60.4 GB 62.3 GB 66.0 GB 73.5 GB 88.6 GB
IQ4_XS (2 parts) Llama-4-Scout-17B_109B 59.9 GB Small 61.4 GB 62.4 GB 64.3 GB 68.0 GB 75.5 GB 90.6 GB
Q4_K_M (2 parts) Mistral-Large-2407_123B_100B 60.0 GB Small 61.6 GB 62.5 GB 64.4 GB 68.1 GB 75.6 GB 90.7 GB
IQ4_XS (2 parts) Command-A-03-2025_111B 60.0 GB Small 61.6 GB 62.5 GB 64.4 GB 68.2 GB 75.7 GB 90.7 GB
IQ4_XS (2 parts) Command-A-03-2025_111B/Fallen-Command-A-111B-v1.1-IQ4_XS 60.0 GB Small 61.6 GB 62.5 GB 64.4 GB 68.2 GB 75.7 GB 90.7 GB
IQ4_XS (2 parts) Command-A-03-2025_111B/TheDrummer_Agatha-111B-v1-IQ4_XS 60.0 GB Small 61.6 GB 62.5 GB 64.4 GB 68.2 GB 75.7 GB 90.7 GB
IQ4_XS (2 parts) Command-A-03-2025_111B/c4ai-command-a-03-2025-IQ4_XS 60.0 GB Small 61.6 GB 62.5 GB 64.4 GB 68.2 GB 75.7 GB 90.7 GB
MXFP4_MOE (2 parts) 63.4 GB - 64.9 GB 65.9 GB 67.7 GB 71.5 GB 79.0 GB 94.1 GB
Q6_K (2 parts) Qwen2-72B 64.3 GB Very small 65.9 GB 66.8 GB 68.7 GB 72.5 GB 80.0 GB 95.0 GB
Q6_K (2 parts) Qwen2-72B/magnum-v2-72b.Q6_K.gguf 64.3 GB Very small 65.9 GB 66.8 GB 68.7 GB 72.5 GB 80.0 GB 95.0 GB
Q6_K (2 parts) Qwen2.5-72B/Qwen2.5-72B-Instruct-Q6_K 64.3 GB Very small 65.9 GB 66.8 GB 68.7 GB 72.5 GB 80.0 GB 95.0 GB
Q4_K_M (2 parts) GLM-4.5-Air_106B 73.5 GB Small 75.0 GB 76.0 GB 77.9 GB 81.6 GB 89.1 GB 104 GB 
Ollama  ollama run hf.co/knifeayumu/LLM_Collection:Q4_K_M
llama.cpp  llama-server -hf knifeayumu/LLM_Collection:Q4_K_M

Generation speed on common GPUs and Macs (estimated)

A rough guide to how fast Q4_K_M (41.4 GB) generates text. Each token reads the model weights once, so the upper limit is memory bandwidth ÷ file size. The guide is 60–80% of that, where most published measurements fall.

Device Memory bandwidth Guide (tokens/s) Estimated Upper limit
Apple M5 Pro 64 GB 307 GB/s 4.4–5.9 7.4
Apple M2 Max 64 GB 400 GB/s 5.8–7.7 9.7
Apple M5 Ultra 96 GB 1,200 GB/s 17–23 29

Estimates only. Prompt processing speed (limited by compute) is not included. Longer contexts and offloading part of the model to the CPU make it slower. Macs are shown only when installed memory is at least 4/3 of what is needed. Devices without published bandwidth (some GeForce cards, etc.) are not shown.

How to run

Example commands using Q4_K_M. To use another quantization, replace the name with one from the table above. How to run local LLMs

Ollama

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

ollama run hf.co/knifeayumu/LLM_Collection:Q4_K_M

llama.cpp

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

llama-server -hf knifeayumu/LLM_Collection:Q4_K_M

LM Studio

Paste knifeayumu/LLM_Collection 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 LLM_Collection (Q4_K_M)?
The file size is 41.4 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 43.9 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Does LLM_Collection run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?
8 GB: up to Q6_K (est. 8.43 GB); 12 GB: up to Q8_0 (est. 12.3 GB); 16 GB: up to Q5_K_M (est. 16.6 GB); 24 GB: up to Q8_0 (est. 24.7 GB) (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run LLM_Collection with Ollama?
Run "ollama run hf.co/knifeayumu/LLM_Collection:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.