Experimental 3.5T K2 merge beats GPT-4.5 and Opus at writing huggingface.co 2 points by bo0tzz 14 hours ago
mynti 14 hours ago "beats at writing" is kind of far fetched since the page provides three generic pormpts with answers from Opus, GPT5 and K3. But a 3.5T model is crazy huge and the answers still feel more or less the same to the other models
"beats at writing" is kind of far fetched since the page provides three generic pormpts with answers from Opus, GPT5 and K3. But a 3.5T model is crazy huge and the answers still feel more or less the same to the other models