Alex Finn Claims Local AI is the Future, 27B Model Now Runs on iPhone
Alex Finn stated that local AI is the future, and if you haven't adopted it yet, it's a delusion.
Today, the 27B model based on Qwen 3.6 (equivalent to Sonnet's level six months ago) was released, enabling mobile operation via PrismML.
Although it is not the smartest or fastest and will not immediately replace major workflows, the trajectory is clear: 24 months ago, local models required million-dollar data centers, and now they are on mobile devices.
In 12 months, the Mac Mini will run a better model, urging people to learn local AI now to avoid being left behind.
Market mechanisms indicate that the progress of local AI boosts demand for consumer hardware and privacy computing, with related chip and optimization technology companies gaining attention, while reliance on centralized AI is expected to decline.
Source: Public Information
ABAB AI Insight
Local AI is rapidly developing, similar to the shift from cloud to edge computing, having achieved breakthroughs in device-side deployment through model compression technology.
In terms of capital pathways, optimization companies like PrismML are shifting resources towards mobile deployment, with developers and users turning to privacy-focused local solutions, while large model companies need to address localized competition strategically.
Similar to the optimization cycles of AI hardware and software, the current phase of local models is accelerating deployment on mobile devices.
Essentially, this represents a paradigm shift in computing, with the 27B model on the iPhone marking a leap of local AI from data centers to consumer devices, driven by model compression and hardware collaboration to enhance privacy, security, and low latency, thereby reshaping AI usage scenarios and capital allocation patterns.
ABAB News · Cognitive Laws
The trajectory of local AI from data centers to mobile is the future direction.
Privacy and offline capabilities will become core competitive advantages.
Ignoring localization means missing the next wave of AI infrastructure.