Artificial analysis index scores Qwen 3.5 27b: 35 Qwen 3.6 27b: 38 Qwen 3.8 27b: 52 What the hell kind of a jump was that? Even if it is benchmaxxed, the jump is insane. Qwen3.6 35b A3b: 32 That's ~6 points behind its dense 27b model, but is ~5x faster for inference given only 3b active params/token
