Not much to say about this model other than that it needs a slightly ridiculous weight to get gemmy results. 1.4 might do it, but may need higher for more complex prompts