SmallThinker-3b is the first reasoning AI model you can run on a phone. Here it is running at 12 tok/s. Performs similar to gpt-4o in a number of domains. Insane that it can run in your hand.
@localghost
-
Running Larger AI Models on M-Series MacBooks with 48GB RAM
By
–
You can run a much larger model than that if you have an m-series MacBook and 48gb
-
Decentralized Personal AI and Cloud-Based Super Intelligence
By
–
both: decentralized ai's on each person and super intelligence in the cloud
-
Using AI for Complex Problem Solving and Synthetic Data Generation
By
–
I talk to it about everything. Hard problems, work, curiosities. Also use it for small batches of synthetic data.
-
o1 Pro Usage Costs Exceed $200 Significantly
By
–
Having used both o1 Pro and the o1 API (as a price reference) I can confidently say I have used > $200 in o1 Pro queries by many multiples.
-
Local Agents: The Next Exciting Frontier in AI Development
By
–
local agents are going to be exciting
-
Local AI Setup: LM Studio and Apollo Integration Guide
By
–
it’s pretty simple, all you need is LM studio on your computer and apollo on your phone
-
LM Studio and Apollo: Local LLM Setup Guide
By
–
it’s pretty simple, all you need is LM studio on your computer and apollo on your phone
-
MacBook Air Now Runs Powerful AI Models for Home Networks
By
–
In 2025 even a MacBook Air can host good AI models to your whole network.
— Aaron Ng (@localghost) 5 janvier 2025
Here’s Ministral 8B streaming from my Pro to my phone. One of the models that would work well on an Air. pic.twitter.com/yqPIguvuk9In 2025 even a MacBook Air can host good AI models to your whole network. Here’s Ministral 8B streaming from my Pro to my phone. One of the models that would work well on an Air.