Nicolas MØUARTHow many items in this cabinet of curiosities?
I counted ~60 items, while Gemini 3 pro -> 50, Gemma4 26B A4B -> 51, and Qwen3.5 35B A3B -> 48.
I think best approach with these tools is to use them as drafters: result has to be checked in any case, yet it can save time, assuming it is not all wrong (it depends on the topic, it's not obvious if the LLM has been trained well, or not).
Transparency matters, whatever the AI industry calls them, they are not really open source.
