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

> Retained reddit record from reddit:LocalLLaMA.

Canonical: https://highsignal.app/data/records/9fa3d1f28d22c309

[← sources](https://highsignal.app/data)


source: [reddit](https://highsignal.app/data/reddit) (reddit:LocalLLaMA)  published 2026-10-03 06:18 ingested 2026-10-04 02:38

[original source ↗](https://www.reddit.com/r/LocalLLaMA/comments/1wwfyv6/qwen4exp_halve_the_indexer_score_memory_by/)

## 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
