llmspeed

Feel how fast a local LLM runs before you buy the hardware.

Pick a GPU or Mac and a model, and watch the answer stream at the real speed, including the wait before the first token.

Features:

• Real-time simulation: text streams at the predicted tokens/sec, after a realistic time to first token for your prompt length
• Race mode: run two setups side by side, e.g. a 5090 against a Mac Studio
• Memory fit: see whether a model fits at each quantization, needs partial offload to system RAM, or won't run at all
• 30 GPUs, Macs and CPUs × 15 popular models, across CUDA, Metal, ROCm, Vulkan and CPU
• Shareable links to any setup or race