They're quite selective in benchmarks, c.f. notably only bad one is 10% on TerminalBench. It's a really addled model, one time I said "Hi" and it built out a 4 panel hello world app with (fake) weather, a todo list, and a couple other things I forgot. I wouldn't be comfortable saying "ignore the #s!" except when I complained it was trash and way overcooked on agentic coding yet not good at it, and a couple DeepMind ML people liked the tweet.
