Model Quant Measured Estimated RAM used App Source Date
๐Ÿ—ฃ๏ธ Kokoro 82M FP16 200 tok/s 200 tok/s +0% 0.5 GB Xybrid CLI editorial 2026-03-02
Llama 3.2 1B Llama 3.2 1B Q8 67 tok/s 37 tok/s +81% 1.8 GB Ollama community 2026-03-01
Phi-4 Mini Phi-4 Mini Q4 48 tok/s 13 tok/s +269% 2.4 GB LM Studio editorial 2026-02-25