VRAM Estimator

Precisely estimate model weights, KV Cache, activations, and CUDA overhead. Supports model presets, GPU presets, and smart optimization suggestions when VRAM is insufficient.

🧠 Model Preset

Select Model Use Custom Parameters Quantization

⚙️ Custom Parameters

Parameters (B) Hidden Size Layers Attention Heads KV Heads (GQA) Max Context

🚀 Runtime Parameters

Context Length Concurrent Requests Max GPU Utilization Recommended 0.85–0.9, do not exceed 0.92 in production

🎮 GPU Configuration

Select GPU GPU Count VRAM per GPU (GB)

Calculating...

-- GB Est. Required

-- GB Available VRAM

0% --% 100%

VRAM Breakdown

Model Weights --

KV Cache --

Activations --

CUDA Overhead --

VRAM Estimator

🧠 Model Preset

⚙️ Custom Parameters

🚀 Runtime Parameters

🎮 GPU Configuration

VRAM Breakdown

Per-GPU Allocation

💡 Optimization Suggestions

VRAM Estimator

🧠 Model Preset

⚙️ Custom Parameters

🚀 Runtime Parameters

🎮 GPU Configuration

VRAM Breakdown

Per-GPU Allocation

💡 Optimization Suggestions

Related Tools