PrismML has developed a novel compression technique that significantly reduces the size of large language models while retaining most of their performance. Their latest model, Bonsai 2.27B, compresses Alibaba's Qwen3.8 27B model by 9x-10x, making it small enough to run on personal computers and potentially smartphones. The company, founded by Caltech researchers, aims to apply this 'ternary' weight compression to even larger models in the future, enabling more accessible and private AI. AI
IMPACT Enables more efficient deployment of advanced AI models on edge devices, potentially increasing accessibility and privacy.
RANK_REASON This is a product release from a startup focused on LLM compression, not a frontier model release from a major AI lab.
- Alibaba Group
- Apple Inc.
- Babak Hassibi
- Berkeley
- Bonsai 2.27B
- California Institute of Technology
- Cerberus Capital Management
- Databricks
- Donostia International Physics Center
- Ion Stoica
- Khosla Ventures
- Multiverse Computing
- PrismML
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
