Qwen 3.8 27B runs locally and ties GPT-5.6 Luna in benchmarks

haider1 · x · 2026-08-31

Alibaba's Qwen 3.8 27B is an open-weight model that runs locally on machines with roughly 24–32GB of memory depending on quantization. The author notes it ties GPT-5.6 Luna and Gemini 3.6 Flash, and sits just 1 point behind GPT-5.6 Terra — arguing this should have shaken OpenAI and Anthropic.

Original post →

More from Models

Models channel →