3 ms·Llama 4 underperforms: a benchmark against coding-centric models5 points by sylvainkalache 2y ago