We tested the performance of Lama 3.1, Lama 3.2, Mistol, Gemma 1, Gemma 2, 53, and Quen 2.5, on a 14-inch MacBook Pro M2 Pro with 48GB of RAM. Our goal was to determine how fast these models could run on this computer and assess their storytelling capabilities.
We used a text prompt to generate a 500-word story about how AI took John’s job as a French to English translator. We analyzed the results for efficiency, emotional depth, and avoidance of melodrama.
Lama 3.2 emerged as the top performer, generating a highly detailed story with strong character development at a blazing-fast speed of 64 tokens per second. Gemma 2 and Quen 2.5 also impressed, delivering compelling stories at respectable speeds.
2 responses to “Benchmarks: Running Large Language Models on a 14-inch MacBook Pro M2 Pro###”
I can’t help but feel that this is the kind of work that will stand the test of time. There’s a timeless quality to your writing that transcends trends and fads. It’s the kind of piece that people will look back on years from now and still find meaning in.
Hello! Thank you so much for this thoughtful comment.