# Show HN: Maple – Sub\-2\-BPW Moe Pruning to Run Gemma 4 26B at 45 TPS on a 6GB GPU

- HackerByte page: <https://hackerbyte.net/en/stories/a-technique-for-improving-the-efficiency-of-large-language-models-llms-on-gpus-50410253>
- Original source: <https://github.com/AmSach/maple>
- Hacker News discussion: <https://news.ycombinator.com/item?id=50410253>
- Submitted to Hacker News: 2026-11-08T03:48:04.000Z

A technique for improving the efficiency of large language models \(LLMs\) on GPUs.

## AI-generated summary

No summary is available in this language yet.