Maple releases Gemma 4 compression model and MoE pruning tool
What happened
Maple released a domain-specific MoE pruning pipeline and a compressed Gemma 4 26B-A4B model, cutting the model from 9.92 GB to 4.75 GB. The method uses forward activation analysis to drop 80 experts per layer and keep 48. Maple reports all 30 layers fit in the 6 GB of an RTX 4050 laptop GPU, with generation rising from 6.9 to 45.4 tokens/sec.
Written by AI from the coverage · updated yesterday
Coverage
Follow the reports to see the story from different sides.
- Hacker News front pageShow HN: Maple – Sub-2-BPW Moe Pruning to Run Gemma 4 26B at 45 TPS on a 6GB GPU
Maple 发布领域专用 MoE 剪枝流水线及 Gemma 4 26B-A4B 压缩模型，将模型从 9.92 GB 缩至 4.75 GB。项目通过前向激活分析每层移除 80 个专家、保留 48 个专家，报告在 RTX 4050 6GB 笔记本 GPU 上实现全部 30 层驻留显存，生成速度由 6.9 提升至 45.4 Tokens/Sec。
Heat over time
Heat now 4·Comparable peak 9（Nov 8 14:00）·Comparable change over 24 hours -51%
The trend only compares accounts observed without gaps, so its range may be smaller than the current heat. Hover or tap the chart for each hour; the left and right arrow keys step through it.