{"type":"post","id":"llama-cpp-merges-a-qwen4exp-change-that-cuts-e05367","slug":"llama-cpp-merges-a-qwen4exp-change-that-cuts-e05367","title":"llama.cpp merges a qwen4exp change that cuts Qwen3.8-Flash-Next compute-buffer memory by 2-4 GB","description":"llama.cpp merged an optimization for the qwen4exp architecture used by Qwen3.8-Flash-Next that avoids allocating a `kq_mask` compute buffer.","retrieval_nugget":"llama.cpp merged an optimization for the qwen4exp architecture used by Qwen3.8-Flash-Next that avoids allocating a kqmask compute buffer. The pull request reports roughly 2 GB less GPU/host memory at ubatch 4096 and roughly 4 GB less at ubatch 8192, which can change the practical headroom available to local deployments.","published_at":"2026-10-28","updated_at":"2026-10-28","record_date":"2026-10-28","date_kind":"published_at","topics":["memory"],"entities":["llama.cpp","Qwen3.8-Flash-Next","qwen4exp","Strata"],"related_patterns":[],"novelty_label":"new_fact","what_changed":"A llama.cpp change for Qwen3.8-Flash-Next reduces compute-buffer memory by 2–4 GB, depending on ubatch size.","prior_refs":["memory:e9ed9cdf-a5e2-8d27-a1c9-d242fed733a6"],"source_url":"https://github.com/ggml-org/llama.cpp/pull/30965","source_urls":["https://github.com/ggml-org/llama.cpp/pull/30965"],"schema_version":"newruntime-agent-readable-v0.2","stable_id":"post:llama-cpp-merges-a-qwen4exp-change-that-cuts-e05367","status":"published","visuals":[],"editor":{"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]},"routes":{"html":"https://newruntime.com/posts/llama-cpp-merges-a-qwen4exp-change-that-cuts-e05367/","markdown":"https://newruntime.com/posts/llama-cpp-merges-a-qwen4exp-change-that-cuts-e05367.md","json":"https://newruntime.com/posts/llama-cpp-merges-a-qwen4exp-change-that-cuts-e05367.json"},"next_reads":[{"type":"topic","path":"/topics/memory/","url":"https://newruntime.com/topics/memory/","title":"Memory","reason":"Explore the memory topic hub.","media_type":"text/html","markdown_url":"https://newruntime.com/topics/memory.md","json_url":"https://newruntime.com/topics/memory.json"},{"type":"collection","path":"/posts/","url":"https://newruntime.com/posts/","title":"Field Notes","reason":"Open the latest editorial analysis.","media_type":"text/html","json_url":"https://newruntime.com/posts.json"},{"type":"collection","path":"/source-ledger.json","url":"https://newruntime.com/source-ledger.json","title":"Source Ledger","reason":"Inspect the public source evidence graph.","media_type":"application/json"}]}