PrismML is a startup that uses mathematical wizardry to reduce the size of large language models, and could be used to bring more advanced AI capabilities to the iPhone in the future.
The company has already managed to compress the 54GB Qwen 3.6 model to just 4GB. And importantly, the technology used does not affect the performance of the model.
Apple is already keen to find ways to shrink LLM, which typically runs on servers. The company hopes to run these models on devices, but has so far been limited by the vast amounts of data required for more advanced models.
Now, according to a report, informationApple has identified PrismML as a potential partner in its AI reduction efforts.
billions of parameters
The Qwen 3.6 model is so large because it has 27 billion parameters, which allows it to perform more advanced tasks than smaller models. Models running on phones typically have billions of parameters and limited functionality.
The Qwen 3.6 model that PrismML has already been built to run on iPhone is reportedly capable of complex chat and inference. You can also power fully autonomous agents.
Apple is already working on shrinking models for the iPhone, but performance has been an issue. However, PrismML’s technology has no such problems, which makes it particularly appealing.
With that in mind, it’s no wonder Apple is taking notice. According to the report, the company has already held a meeting with PrismML to discuss how to use its technology to bring more advanced AI technology to the iPhone.
After all, there is no guarantee that the two companies will be able to work together. Even if that happens, it’s unclear when iPhone and Siri AI owners will be able to benefit from this partnership.
