Hacker News new | ask | show | jobs
by davely 313 days ago
Not entirely unexpected given that these models have 2x - 5x more parameters than the 120b gpt-oss model.

It’s impressive their performance gets so close given the size!