Swift-1.5-Qwen3.8-27B in hipfire for AMD cards: ~2.2x faster end-to-end than Q4_K_S GGUF on a 7900XT

🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1xcwkfi/swift15qwen3827b_in_hipfire_for_amd_cards_22x/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-10-23 06:17:24
Reddit
From the LocalLLaMA community on Reddit: Swift-1.5-Qwen3.8-27B in hipfire for AMD cards: ~2.2x faster end-to-end than Q4_K_S GGUF…
From the LocalLLaMA community on Reddit: Swift-1.5-Qwen3.8-27B in hipfire for AMD cards: ~2.2x faster end-to-end than Q4_K_S GGUF…
Explore this post and more from the LocalLLaMA community