Qwen 27B (IQ4_XS, Q8/Q4 KV) at full 262k context on a 16GB RX 9070 XT: 80+ t/s decode with Adaptive KV Streaming

🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1x89ptz/qwen_27b_iq4_xs_q8q4_kv_at_full_262k_context_on_a/
🏷️ 分组: Reddit｜AI·自托管·VPS
🕒 时间: 2026-10-17 21:01:59
Reddit
From the LocalLLaMA community on Reddit: Qwen 27B (IQ4_XS, Q8/Q4 KV) at full 262k context on a 16GB RX 9070 XT: 80+ t/s decode…
From the LocalLLaMA community on Reddit: Qwen 27B (IQ4_XS, Q8/Q4 KV) at full 262k context on a 16GB RX 9070 XT: 80+ t/s decode…
Explore this post and more from the LocalLLaMA community