Model Quant Measured Estimated RAM used App Source Date
Phi-4 Mini Phi-4 Mini Q5 105 tok/s 35 tok/s +200% 2.9 GB Ollama editorial 2026-02-15
Llama 3.2 3B Llama 3.2 3B Q8 78 tok/s 41 tok/s +90% 3.8 GB LM Studio editorial 2026-02-18
Mistral 7B Mistral 7B Q4 71 tok/s 19 tok/s +274% 4.3 GB LM Studio editorial 2026-02-15
Qwen 2.5 7B Qwen 2.5 7B Q4 64 tok/s 18 tok/s +256% 4.6 GB Ollama community 2026-03-01
Gemma 3 12B Gemma 3 12B Q4 40 tok/s 11 tok/s +264% 7.3 GB LM Studio editorial 2026-03-05