Local AI is moving too fast now, there is something new every day.
Qwen 3.8 27B running on consumer laptops and beating cloud models people pay for. A new engine called FreeToken putting a 35B model on an 8GB laptop. 128GB local AI boxes from AMD and NVIDIA becoming normal desk
AI is getting more capable, but it’s also getting smaller, more local, and more collaborative.
🧠🤖 We tested Qwen 3.8, local models, multi-agent workspaces, new coding tools, and more to see where AI workflows are actually heading. The bigger question we kept coming back to:
Qwen4-Max OUTPERFORMED Fable 5 and GPT-5.6 Sol
Already on Arena as "paloma" this is an Opus 5-level model on the frontend 3 damn weeks since Qwen 3.8, which is significantly inferior to Opus 5
A single 3090 running Qwen 3.8 27b at 220 t/s + 150k context
Q8 on weights and KV cache Let the bastard think 😈
You have reached the end of the archive
All of qwen38