Can you spot the difference?
👀 Speculative decoding helps accelerate generative AI at the edge, especially on NVIDIA Jetson. Qwen 3.8 27B went from 13 to 35 tokens/sec, while Nemotron 3.5 Lightning went from 65 to 115. See how it works:
Not bad.. pi + Qwen 3.8 27b running locally w/ threejs prompt shared below 👇
Anyone have tips for managing context? best harness?
They told us qwen 3.8-27b is literally “claude opus 4.6 on your local pc for free”...
Bro don’t be delulu everyone on my timeline is like “open source won, closed models are going to be dead soon” but nobody talks about how cooked the actual setup is: 1. the “17gb vram” claim
Gemini 3.7 Flash High vs Claude Opus 5 High vs Grok 4.6 High vs Qwen 3.8 27B
3.7 Flash created it in under a min (name is justified) I didn’t expect Qwen 3.8 27B to be this good for its size detailed 3D Clockwork Helicopter using Three.js
You have reached the end of the archive
All of qwen38