How much memory do I need to run karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 (Q4_K_M)?

The file size is 28.4 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 30.1 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.

Does karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 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: up to Q3_K (est. 24.2 GB) (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).

What is the command to run karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 with Ollama?

Run "ollama run hf.co/RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.

RichardErkhov

karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1

Parameters: 46.7B
Max context length: 32,768
Architecture: llama

Memory needed to run Q4_K_M at 8K tokens

30.1 GB Estimated

File size 28.4 GB Measured

VRAM it fits in (rough guide)

8 GB does not fit
12 GB does not fit
16 GB does not fit
24 GB does not fit
32 GB fits
48 GB fits

See devices and models for 32 GB

The smallest, Q2_K, needs an estimated 19.0 GB at 8K.

Macs cap how much memory the GPU can use, so leave headroom in unified memory.

Data as of Nov 11, 2026

About karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1

karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 is a model published in GGUF format. GGUF files are published by RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf. It has about 46.7B parameters, the llama architecture, a maximum context length of 32,768 tokens (values written in the GGUF).

There are 22 quantizations, with file sizes from 17.3 GB (Q2_K) to 49.6 GB (Q8_0) (actual sizes on Hugging Face).

Running Q4_K_M (28.4 GB) with an 8,192-token context needs an estimated 30.1 GB of memory, and an estimated 33.3 GB at 32,768 tokens. For 8,192 tokens, a GPU with 32 GB or more of VRAM (such as GeForce RTX 5090), or a Mac with 32 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, Q2_K (17.3 GB), needs an estimated 19.0 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

RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf

Published by the original developer About this label

Last checked: Nov 11, 2026

Architecture: llama; Supported before llama.cpp b4409
Parameters: 46.7B
Max context length: 32,768

File size and memory per quantization for karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 (RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf)
Quantization	File size Measured	Quality guide	4,096	8,192	16,384	32,768 max
Q2_K	17.3 GB	Very large	18.5 GB	19.0 GB	20.1 GB	22.2 GB
IQ3_XS	19.4 GB	Large	20.5 GB	21.0 GB	22.1 GB	24.2 GB
IQ3_S	20.4 GB	Medium	21.6 GB	22.1 GB	23.2 GB	25.3 GB
Q3_K_S	20.4 GB	Large	21.6 GB	22.1 GB	23.2 GB	25.3 GB
IQ3_M	21.4 GB	Medium	22.6 GB	23.1 GB	24.2 GB	26.3 GB
Q3_K	22.5 GB	-	23.7 GB	24.2 GB	25.3 GB	27.4 GB
Q3_K_M	22.5 GB	Medium	23.7 GB	24.2 GB	25.3 GB	27.4 GB
Q3_K_L	24.2 GB	Medium	25.3 GB	25.8 GB	26.9 GB	29.1 GB
IQ4_XS	25.4 GB	Small	26.5 GB	27.1 GB	28.1 GB	30.3 GB
Q4_0	26.4 GB	Medium	27.6 GB	28.1 GB	29.2 GB	31.3 GB
IQ4_NL	26.7 GB	Small	27.9 GB	28.4 GB	29.5 GB	31.6 GB
Q4_K_S	26.7 GB	Medium	27.9 GB	28.4 GB	29.5 GB	31.6 GB
Q4_K	28.4 GB	-	29.6 GB	30.1 GB	31.2 GB	33.3 GB
Q4_K_M	28.4 GB	Small	29.6 GB	30.1 GB	31.2 GB	33.3 GB
Q4_1	29.3 GB	Medium	30.5 GB	31.0 GB	32.1 GB	34.2 GB
Q5_0	32.2 GB	Small	33.4 GB	33.9 GB	35.0 GB	37.1 GB
Q5_K_S	32.2 GB	Small	33.4 GB	33.9 GB	35.0 GB	37.1 GB
Q5_K	33.2 GB	-	34.4 GB	34.9 GB	36.0 GB	38.1 GB
Q5_K_M	33.2 GB	Very small	34.4 GB	34.9 GB	36.0 GB	38.1 GB
Q5_1	35.1 GB	Small	36.3 GB	36.8 GB	37.9 GB	40.0 GB
Q6_K	38.4 GB	Very small	39.5 GB	40.1 GB	41.1 GB	43.3 GB
Q8_0 (2 parts)	49.6 GB	Nearly identical	50.8 GB	51.3 GB	52.4 GB	54.5 GB

Ollama  ollama run hf.co/RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf:Q4_K_M

llama.cpp  llama-server -hf RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf:Q4_K_M

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/RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf:Q4_K_M

llama.cpp

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

llama-server -hf RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf:Q4_K_M

LM Studio

Paste RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-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 karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 (Q4_K_M)?: The file size is 28.4 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 30.1 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Does karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 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: up to Q3_K (est. 24.2 GB) (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1 with Ollama?: Run "ollama run hf.co/RichardErkhov/karakuri-ai_-_karakuri-lm-8x7b-instruct-v0.1-gguf:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.