Apple is in talks with PrismML to compress AI models like Qwen 3.6 for the iPhone
Apple is in talks with startup PrismML about a technology that significantly shrinks large language models without losing functionality, according to The Information. The talks began after PrismML compressed Alibaba's open 27-billion-parameter Qwen 3.6 model to run locally on the iPhone 17 Pro. To do it, PrismML changed the usual approach to "sparse architecture."
AI-processed from 3DNews AI; edited by Hamidun News
Apple is in talks with startup PrismML about possibly using its technology for compressing large language models. The technology would let powerful AI models run directly on the iPhone without a cloud connection, The Information reported, citing its sources.
What PrismML has already compressed
The talks began after PrismML achieved a technical breakthrough: the startup compressed Alibaba's open Qwen 3.6 model to run locally on the iPhone 17 Pro.
- Alibaba's Qwen 3.6 model has 27 billion parameters
- PrismML compressed it to run on the iPhone 17 Pro without losing functionality or intelligence
- The technology makes it possible to use powerful AI models without a cloud connection
- The Information reported the Apple–PrismML talks, citing sources
How PrismML tackles overheating
Fitting a model with this many parameters on a smartphone usually requires a "sparse architecture": the device activates only part of the AI's "brain" at any given moment to avoid overheating. PrismML reportedly overhauled this approach to model compression entirely.
What it means
If a deal goes through, Apple will gain a way to run genuinely large AI models directly on the iPhone without the cloud. In recent years, the company has bet on private AI that works without sending data off the device. A partnership with PrismML fits that strategy and could be a notable step toward more powerful and faster on-device AI features.
FAQ
What is PrismML?
It's a startup that has developed technology to significantly compress large language models without losing their functionality or intelligence. According to The Information, Apple is in talks with the company.
Which model has PrismML already compressed?
Alibaba's open Qwen 3.6 model with 27 billion parameters. PrismML compressed it to run locally on the iPhone 17 Pro.
Which iPhone has already run the compressed model?
The Information's report names the iPhone 17 Pro. That's the device on which PrismML ran a compressed version of Qwen 3.6 locally.
- Oct 29, 2026Claude Code Edits Video From a Transcript: How the video-use Skill With 30 000 Stars Works
- Oct 29, 2026AI Verified Part of Signal's Code with Math in Half a Day for $500, Versus Five Months by Hand
- Oct 29, 2026Liquid AI releases LFM3: six multimodal 2B models with a 256K context window
- Oct 29, 20268B-Parameter Agent Operates iOS Apps at Gemini 3.1 Pro Level
Key Questions Answered
What is PrismML?
It's a startup that has developed technology to significantly compress large language models without losing their functionality or intelligence. According to The Information, Apple is in talks with the company.
Which model has PrismML already compressed?
Alibaba's open Qwen 3.6 model with 27 billion parameters. PrismML compressed it to run locally on the iPhone 17 Pro.
Which iPhone has already run the compressed model?
The Information's report names the iPhone 17 Pro. That's the device on which PrismML ran a compressed version of Qwen 3.6 locally.
Ship AI Faster with Founders, Engineers & Weekly Architecture Reviews
Inside Vibe Club (hamidun.com/club) — weekly live masterminds with Zhemal Khamidun, Tech Radar, full LMS access, and 370+ production Claude Code skills.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.