Tagged with

2 articles found

Swift-Qwen3.8-27B: The “Stop Overthinking” Fix That Actually Works

UkisAI’s post-trained Qwen cuts thinking tokens by 58% and doubles speed with under 1% accuracy loss. Here’s how they did it without crippling the model.

#LLM Efficiency#Model Optimization#on-policy distillation...

Edge AI

800 Million Parameters, One Demonslayer: The Sub-1B Model Running Doom on Your Wrist

How Qwen 3.5 0.8B manages complex spatial reasoning and action execution on smartwatch-grade hardware, and what it means for the future of edge AI.

#Edge AI#Model Optimization#On-Device Inference...