Skip to content
localllm
日本語

RichardErkhov

rinna_-_nekomata-7b-instruction

Parameters
7.7B
Max context length
32,768
Architecture
qwen

About rinna_-_nekomata-7b-instruction

rinna_-_nekomata-7b-instruction is a model published in GGUF format. GGUF files are published by RichardErkhov/rinna_-_nekomata-7b-instruction-gguf. It has about 7.7B parameters, the qwen architecture, a maximum context length of 32,768 tokens (values written in the GGUF).

There are 22 quantizations, with file sizes from 3.05 GB (Q2_K) to 8.21 GB (Q8_0) (actual sizes on Hugging Face).

Running Q4_K_M (4.90 GB) with an 8,192-token context needs an estimated 10.3 GB of memory, and an estimated 23.2 GB at 32,768 tokens. For 8,192 tokens, a GPU with 12 GB or more of VRAM (such as GeForce RTX 3080 Ti, GeForce RTX 5070), or a Mac with 12 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 (3.05 GB), needs an estimated 8.42 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/rinna_-_nekomata-7b-instruction-gguf

Published by the original developer About this label

Last checked: Nov 11, 2026

Architecture
qwen
Supported before llama.cpp b4409
Parameters
7.7B
Max context length
32,768
File size and memory per quantization for rinna_-_nekomata-7b-instruction (RichardErkhov/rinna_-_nekomata-7b-instruction-gguf)
Quantization File size Measured Quality guide 4,096 8,192 16,384 32,768 max
Q2_K 3.05 GB Very large 6.27 GB 8.42 GB 12.7 GB 21.3 GB
IQ3_XS 3.47 GB Large 6.69 GB 8.84 GB 13.1 GB 21.7 GB
IQ3_S 3.57 GB Medium 6.79 GB 8.94 GB 13.2 GB 21.8 GB
Q3_K_S 3.57 GB Large 6.79 GB 8.94 GB 13.2 GB 21.8 GB
IQ3_M 3.88 GB Medium 7.10 GB 9.25 GB 13.5 GB 22.1 GB
Q3_K 4.06 GB - 7.28 GB 9.43 GB 13.7 GB 22.3 GB
Q3_K_M 4.06 GB Medium 7.28 GB 9.43 GB 13.7 GB 22.3 GB
Q3_K_L 4.29 GB Medium 7.51 GB 9.66 GB 14.0 GB 22.5 GB
IQ4_XS 4.32 GB Small 7.54 GB 9.69 GB 14.0 GB 22.6 GB
Q4_0 4.51 GB Medium 7.73 GB 9.88 GB 14.2 GB 22.8 GB
IQ4_NL 4.53 GB Small 7.76 GB 9.90 GB 14.2 GB 22.8 GB
Q4_K_S 4.53 GB Medium 7.76 GB 9.90 GB 14.2 GB 22.8 GB
Q4_K 4.90 GB - 8.12 GB 10.3 GB 14.6 GB 23.2 GB
Q4_K_M 4.90 GB Small 8.12 GB 10.3 GB 14.6 GB 23.2 GB
Q4_1 4.96 GB Medium 8.18 GB 10.3 GB 14.6 GB 23.2 GB
Q5_0 5.40 GB Small 8.62 GB 10.8 GB 15.1 GB 23.7 GB
Q5_K_S 5.40 GB Small 8.62 GB 10.8 GB 15.1 GB 23.7 GB
Q5_K 5.71 GB - 8.93 GB 11.1 GB 15.4 GB 24.0 GB
Q5_K_M 5.71 GB Very small 8.93 GB 11.1 GB 15.4 GB 24.0 GB
Q5_1 5.84 GB Small 9.06 GB 11.2 GB 15.5 GB 24.1 GB
Q6_K 6.34 GB Very small 9.56 GB 11.7 GB 16.0 GB 24.6 GB
Q8_0 8.21 GB Nearly identical 11.4 GB 13.6 GB 17.9 GB 26.5 GB 
Ollama  ollama run hf.co/RichardErkhov/rinna_-_nekomata-7b-instruction-gguf:Q4_K_M
llama.cpp  llama-server -hf RichardErkhov/rinna_-_nekomata-7b-instruction-gguf:Q4_K_M

Generation speed on common GPUs and Macs (estimated)

A rough guide to how fast Q4_K_M (4.90 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
Radeon RX 7600 XT 16 GB 288 GB/s 35–47 59
Radeon RX 6950 XT 16 GB 576 GB/s 71–94 118
GeForce RTX 5090 32 GB 1,792 GB/s 219–293 366
Apple M2 16 GB 100 GB/s 12–16 20
Apple M2 Pro 16 GB 200 GB/s 24–33 41
Apple M5 Ultra 96 GB 1,200 GB/s 147–196 245

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/RichardErkhov/rinna_-_nekomata-7b-instruction-gguf:Q4_K_M

llama.cpp

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

llama-server -hf RichardErkhov/rinna_-_nekomata-7b-instruction-gguf:Q4_K_M

LM Studio

Paste RichardErkhov/rinna_-_nekomata-7b-instruction-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 rinna_-_nekomata-7b-instruction (Q4_K_M)?
The file size is 4.90 GB (measured). With a 8,192-token context, the estimated memory including the KV cache is 10.3 GB. The formula is published at https://localllm.az-arte.com/en/guide/memory-calculation.
Does rinna_-_nekomata-7b-instruction run in 8 GB, 12 GB, 16 GB or 24 GB of VRAM?
8 GB: up to Q2_K (est. 8.42 GB); 12 GB: up to Q6_K (est. 11.7 GB); 16 GB: up to Q8_0 (est. 13.6 GB); 24 GB: up to Q8_0 (est. 13.6 GB) (estimated at 8,192 tokens with an f16 KV cache, files from the most-downloaded source).
What is the command to run rinna_-_nekomata-7b-instruction with Ollama?
Run "ollama run hf.co/RichardErkhov/rinna_-_nekomata-7b-instruction-gguf:Q4_K_M" to use the GGUF on Hugging Face directly. To use another quantization, replace the name at the end.