Best Qwen3.8-27B Config for 48GB VRAM
Valuable-Run2129 · reddit · 2026-08-15
User seeks the highest quality setup to run Qwen3.8-27B on 48GB VRAM. Requirements include full context support and offloading prefixes to RAM/storage to minimize reprocessing delays, aiming for optimal intelligence within these constraints.
More from Infra
- llama.cpp adds support for Moonshot Kimi-K3 text model — pmttyji · 2026-08-15
- Self-Hosting Qwen3.8-27B NVFP4 on SGLang Hits 200+ Tokens/Sec — gnukeith · 2026-08-15
- AMD's internal silicon photonics group and Ayar Labs collaboration may surpass MediaTek — zephyr_z9 · 2026-08-15
- Nvidia Cuts OpenAI Data Center Investment to $120B; Anthropic Revenue Surges — The Decoder · 2026-08-15
- Light Trace Photonics Unveils Detachable Photonic Interconnect for AI Data Centers — BenBajarin · 2026-08-15
- Troubleshooting vLLM on AMD v620 for Qwen Models — Thin_Pollution8843 · 2026-08-15