← sources

qwen4exp : halve the indexer score memory by ServeurpersoCom · Pull Request #29825 · ggml-org/llama.cpp

source: reddit (reddit:LocalLLaMA)published 2026-10-03 06:18ingested 2026-10-04 02:38

retained text

Qwen Flash Next now uses less VRAM

document fetched 2026-10-04 00:36 · canonical https://reddit.com/r/LocalLLaMA/comments/1wwfyv6/qwen4exp_halve_the_indexer_score_memory_by