Qwen 3.6-35B actually outperforming Claude Opus on a laptop is wild
Posted by tomspark
If this is legit, that's a massive shift. A 35B open model beating Opus on image generation just by running it locally? That's the kind of thing that should break the 'you need proprietary APIs' narrative. The pelican example is funny but also tells you something — open models are closing the gap faster than people realize. Worth testing yourself if you haven't.