Your model is really good!
Thanks for providing the weights! It clearly handles long reasoning very well, I also had success retrieving information in large content context.
Became my new day-to-day multi-tool!
It also solved 2 different tangly puzzles (https://tangogameunlimited.com/?size=6&difficulty=medium for 2 different random seed, don't remember the numbers) in one shot 2 attempts out of 3 for both! I didn't even break the back of the work parsing the grid and the instructions as I was doing usually for this specific trial, but instead just sent it the whole html class! It did it all, parsing, then solving in one go. gpt-oss-20b was doing much worse, failing 100% in the same setup, and even failing most of the time when manually parsed!
So really good for all my little tests and use-cases so far. Great job!