Initial benchmarks on my M1 Pro, about 10x faster at describing screenshots than qwen3.5-4b. If the quality is comparably, as the model sheet implies, that’s in

Initial benchmarks on my M1 Pro, about 10x faster at describing screenshots than qwen3.5-4b. If the quality is comparably, as the model sheet implies, that’s incredible. But it seems this isn’t a reasoning model?