Qwen3.8-27B-GGUF

Overview

Qwen's new 27B flagship dense model - native vision, flexible thinking control with reasoning_effort, and a big jump on agentic coding benchmarks over 3.6. The bar-setter for a 24 GB card.

Good for

• agentic
• chat
• coding

Recommended hardware

≈ 24 GB VRAM to run this 27B model at the UD-Q5_K_XL quantization.

Fits 24 GB+ consumer GPUs (e.g. RTX 3090, RTX 4090)

Links

• View on HuggingFace
• Run on Infersec