Analyzing your hardware configuration to determine the most capable model you can run locally with good performance...
You
Entry (1B)Mid (7B)Large (70B)SOTA (1T+)
—%
away from top-tier SOTA
Model Tier Comparison
Model Tier
Example Models
Params
RAM Needed
Your Status
Note: These estimates assume quantized models (4-bit) for consumer hardware. Your actual performance depends on quantization quality, model architecture, and optimization tools (llama.cpp, vLLM, etc.). SOTA models like GPT-4o, Claude 3.5, and Gemini Ultra are estimated at 1T+ parameters.