You can now run a legit 27B parameter model completely offline on a normal consumer PC🚀
Qwen 3.8 27B. Download LM Studio (free) → search “Qwen 3.8” → official version from the Qwen team on Hugging Face. They have 4-bit, 5-bit, 6-bit and 8-bit quantizations. Higher bits =
DeepSeek V4 Flash • Qwen 3.8 Flash
GLM 5.3 Flash • Gemini 3.7 Flash
262K context. On a 16GB RTX 5070 Ti.
🤯 Qwen 3.8 27B Q3 hits ~25 tok/s while an adaptive llama.cpp fork streams KV cache between RAM ↔ VRAM. Stock llama.cpp starts thrashing around ~120K context. Same consumer GPU. 2x+ the usable context. This could be huge for local LLMs.
The underlying Qwen 3.8 Max model is huge
2.4 trillion total parameters. But only around 95 billion are activated for an individual task. That mixture-of-experts approach lets it pull in the relevant parts of the model instead of activating everything.
You have reached the end of the archive
All of qwen38