Uploaded August 2026 | Updated September 2026, 1 week ago

---

Qwen 3.8 27B is small enough to make local AI practical on increasingly attainable hardware — but its performance is starting to overlap with models in a very different weight class.

Corey Noles and Grant Harvey break down what “27B” actually means, why quantization matters when you run an LLM locally, the VRAM and context limitations Corey hit in real-world testing, and why Qwen’s position on current AI model benchmarks is so interesting.

We also look at the increasingly crowded frontier model race and why benchmarks alone don’t always tell you which AI model is actually best for your work.

Subscribe to Gadget Hacks: bit.ly/GadgetHacksSub

Follow Gadget Hacks on:
Facebook: facebook.com/gadgethacks
Twitter: twitter.com/GadgetHax
Flipboard: https://flip.it/mF4IuP
Threads: threads.net/@gadgethax

0:00 Analyzing model performance
2:58 Intelligence index review
5:41 Newcomer models and usage
8:15 Historical context and value