LeapQuant: training-free 4/8-bit linear attention quantization, 4.25x faster kernels
berkeley_ai · x · 2026-11-13
Linear attention is slow because the recurrent state is read and rewritten every token. LeapQuant is the first training-free method enabling 8-bit or even 4-bit recurrent states: across Qwen, Kimi, and GLM models it matches FP32 accuracy, speeds linear attention kernels by 4.25x, boosts inference throughput up to 2x, and cuts memory by up to 56%. Ilya Sutskever amplified the work.
Related event: LeapQuant: Training-Free Quantization Speeds Up Linear Attention(4 posts)→
More from Research
- Study: AI startups raise 151% more in North America but no premium in Africa — danielrock · 2026-11-13
- MIT's AI copilot boosts structural buckling resistance 39% with only 15 seconds added — Dr_Singularity · 2026-11-13
- Robopedia launches Xperience 2, a 'gold standard' Physical AI dataset — liuziwei7 · 2026-11-13
- Zhejiang U & Baidu's Agent-R²: 4B agent beats larger open-source baselines on GDPval — ZhejiangUniversity · 2026-11-13
- BirdDreams driving world model generates perspective + BEV video from a single image — lealtaixe · 2026-11-13
- Ladish: AI takeover isn't magic, just scaling pretraining + RL further — JeffLadish · 2026-11-13