I plugged Qwen 3.8 27b into DeepSeek Harness and asked for a replica Age of Empires 2

Returned to my Studio 70 minutes later to find the Mac still purring... So far my local model tests on this prompt have produced unsharable results Let's see what we get

bartslodyczkaAug 241

Qwen 3.8 27B Q4 quant destroyed Q8 at voxel island creation ⛏️

Fede(URU) 🇺🇾Aug 24

Is 0x Alpha actually a public test for someone’s unreleased frontier model?

Same coding test: 0x Alpha: 1.2M tokens · FREE Grok 4.6: 2–4M tokens · $4.20 Qwen 3.8: 550K tokens · $1.90 Qwen built the best result. Grok was buggy. 0x Alpha was a little better — and cost

noclipepeAug 2411

I ran Qwen 3.8 27B on a single 8GB GPU.

IQ4_XS + Q4 KV cache at 70k context. The token speed hovers around 4-5 t/s. Slow but it's still amazing that you could get a result that can outperform even Claude Opus 4.6 on some of my test. All in one-shot with zero agent looping.

Red StaplerAug 242