A Hacker News thread with 1,500+ comments reveals Qwen 3.6 35B-A3B dominates local model mentions at 33%, followed by the 27B variant at 20%. Pi leads agent harnesses at 49%. Users report a 5x speedup versus Claude Opus's 15x, but value privacy, zero cost, and offline capability.
Tap to vote and see what everyone thinks.
Summary by ByteBrief
Google's DiffusionGemma boosts speed 4x