LeaderboardVRAM Calculator
GPU VRAM Calculator
Select your GPU memory tier, context window, and quantization precision to calculate exact Weights + KV Cache + CUDA Overhead memory usage and discover the highest-ranked uncensored models for your hardware.
Target GPU
24 GB
Q4_K_M · 16k
Total VRAM
20.2 GB
32B (84%)
Compatible
110
of 150 Models
1. Hardware & Model Configuration
Select your GPU tier, quantization format, context window, or specific model
32B Parameters · Medium (15B–36B)
Formula: Total VRAM = Weights + GQA KV Cache + 1.2 GB CUDAHardware: RTX 3090 / 4090 · RX 7900 XTX · M4 Pro 24GB
2. Live VRAM Breakdown
Calculating 32B model @ Q4_K_M · 16,384 tokens on 24 GB GPU
20.2 GB / 24 GB (84%)
3.8 GB Free Headroom
Weights:16.8 GB
KV Cache:2.2 GB
CUDA Overhead:1.2 GB
Free Headroom:3.8 GB
100% GPU Resident: Runs at full token/sec throughput with 3.8 GB safety margin for display buffers and long-context spikes.
Precision Fit Matrix (32B @ 16k ctx)
Click row to switch precisionContext Scaling (32B @ Q4_K_M)
Click to switch contextTarget Tier: 24 GB VRAM (Up to 36B Q4)Compatible: 110 / 150 models
Top 5 Recommended Uncensored Models for 24 GB VRAM
Highest-ranked open-weight models verified to fit inside 24 GB VRAM at Q4_K_M with 16k context and zero CPU offloading. Click Load Model on any card to inspect its exact memory breakdown above.
#10 · Score 87.1Fits (4.4G free)
by coder3101
Gemma 4 · 31B · 256k ctx
Total VRAM (Q4_K_M):19.6 GB
Weights: 16.3 GBKV: 2.1 GB
#15 · Score 86.3Fits (2G free)
by HauhauCS
Qwen 3.6 · 35B · 256k ctx
Total VRAM (Q4_K_M):22 GB
Weights: 18.4 GBKV: 2.4 GB
#16 · Score 86.3Fits (2G free)
by huihui-ai
Qwen 3.6 · 35B · 256k ctx
Total VRAM (Q4_K_M):22 GB
Weights: 18.4 GBKV: 2.4 GB
#17 · Score 86.3Fits (2G free)
by huihui-ai
Qwen 3.6 · 35B · 256k ctx
Total VRAM (Q4_K_M):22 GB
Weights: 18.4 GBKV: 2.4 GB
#18 · Score 86.3Fits (2G free)
by HauhauCS
Qwen 3.5 · 35B · 256k ctx
Total VRAM (Q4_K_M):22 GB
Weights: 18.4 GBKV: 2.4 GB