AI BriefingPrismML Compresses 27B Model for iPhone
·2026.07.13 16:59
Key point
PrismML unveils technology that compresses the Qwen-3.6-27B model to under 4GB to run on iPhone.
Details
Caltech spinoff startup PrismML announced that it has succeeded in compressing Alibaba's open-source model Qwen-3.6-27B so that it can run on the iPhone 17 Pro.
The key technical features are as follows:
- Model Compression: The model size, previously about 54GB, has been dramatically reduced to under 4GB.
- Performance Retention: Unlike typical compression methods, it minimizes performance degradation while still enabling complex chat, reasoning, autonomous agent, and software coding tasks.
- On-device AI Acceleration: PrismML reduced the model size through mathematical techniques, suggesting that a significant portion of future AI computation will be handled on-device rather than in the cloud.
The open-source model is scheduled to be released for download this coming Tuesday.
This summary was generated automatically by AI. Check the original for the author's claims and context. Copyright belongs to the original author.
Our guide explains how the AI works. Report summary errors, attribution issues, or removal requests via Contact.