Qwen 3.8 27B Q4_K_M - 90 tokens/sec on a single NVIDIA RTX 4090 (24 GB VRAM) with Dflash2!
(MTP 60 tps -> 90 tps Dflash2!!!!) Local AI moves so fast (literally!) it’s terrifying. Z lab just dropped DFlash 2 for Qwen 3.8 27b and Muse Glimmer. I patched llama.cpp (PR #27342) and
I asked Qwen 3.8 Uncensored to create the Silicon Valley Show but for 2026 with Bytedance Seedance 2.5 / Google Omni.
AI Slop big time.
Built the perplexity comet clone with qwen 3.8
Qwen 3.8 Max just beat Fable 5 and GPT-5.6 Sol Max at using a computer.
86.1 on OSWorld Verified versus 85.0 and 83.2. Alibaba ran it autonomously for 16 days straight.
You have reached the end of the archive
All of qwen38