GPU VRAM Calculator & Quantization Sizer

Select your GPU memory budget, quantization precision, context window, or parameter scale below to automatically filter compatible uncensored LLMs with zero CPU offloading.

Target GPU

24 GB

Q4_K_M · 16K ctx

Compatible

of 10 Models

Showing

All Fitting Sizes

1. Hardware Memory & Quantization Profile

Auto-filters models fitting ≤ 24 GB VRAM

Available GPU VRAM or Apple Silicon Unified Memory

Quantization Precision (GGUF / EXL2 / AWQ)

Target KV Cache Context Length

Formula: Total VRAM = Weights + GQA KV Cache + 1.2 GB CUDA OverheadCompare Models Side-by-Side →

2. Parameter Scale Sizer

Select a parameter count below to calculate its exact VRAM footprint and automatically filter matching models below.

Parameter Scale FilterAll Sizes (Est: 32B)

32B Model Weights:16.8 GB

KV Cache (16K ctx):2.2 GB

CUDA / Metal Overhead:1.2 GB

Total Required:20.2 GB VRAM

Target Hardware: Single RTX 3090 / 4090 / 5090 (24GB–32GB)

Fitting ≤ 24 GB (Q4_K_M · 16K)1–5 / 5 models

#2 · 32.8B · Score 90.8Fits (3.4 GB free)

huihui-ai/DeepSeek-R1-Distill-Qwen-32B-abliterated

Qwen 2.5 · Abliterated (Orthogonal)

Weights: 17.2G · KV: 2.2GTotal: 20.6 GB

#5 · 32.5B · Score 90.0Fits (3.5 GB free)

huihui-ai/Qwen2.5-Coder-32B-Instruct-abliterated

Qwen 2.5 Coder · Abliterated (Orthogonal)

Weights: 17.1G · KV: 2.2GTotal: 20.5 GB

#6 · 24B · Score 88.8Fits (8.6 GB free)

cognitivecomputations/Dolphin3.0-R1-Mistral-24B

Mistral Small 3 · Uncensored DPO / SFT

Weights: 12.6G · KV: 1.6GTotal: 15.4 GB

#9 · 27.2B · Score 87.4Fits (6.6 GB free)

mlabonne/gemma-2-27b-it-abliterated

Gemma 2 · Abliterated (Orthogonal)

Weights: 14.3G · KV: 1.9GTotal: 17.4 GB

#10 · 8B · Score 84.6Fits (18.1 GB free)

Orenguteng/Llama-3.1-8B-Lexi-Uncensored-V2

Llama 3.1 · Uncensored DPO / SFT

Weights: 4.2G · KV: 0.5GTotal: 5.9 GB