OpenAI's latest model, o3, has achieved unprecedented performance on benchmarks like ARC-AGI, sparking debates about the dawn of artificial general intelligence (AGI).

I’m excited to introduce Pocket — the app that brings powerful AI to your iPhone, entirely offline. With Pocket, you can run advanced AI models on your device, keeping your data private and secure.
apps.apple.com/de/app/pocke...

thats very cool! but ollama can't be compared with native framework like MLX which use gpu acceleration. thus, comparing the performance would be nonsense

Look no internet at all!

Running AI locally isn't for everyone. It requires:
Hardware Resources: High-end GPUs or specialized accelerators may be needed for performance.
Setup Time: Initial setup and optimization can be time-consuming.
Maintenance: Ongoing updates and troubleshooting are your responsibility.

8. Experimentation and Learning
Hands-On Experience: Hosting AI locally is a great way to learn more about machine learning and neural networks.
Control Over Updates: You can experiment with new architectures or models without waiting for external providers to update their offerings.

7. Independence from Providers
No Vendor Lock-in: By running AI locally, you avoid becoming dependent on a specific provider's ecosystem, which could change pricing, policies, or availability over time.
Local AI bypasses this problem.

6. Transparency
Understandable Behavior: With local AI, you can inspect and modify the model's architecture or weights, giving you insights into its workings.
Open Source Benefits: Many local models are open-source, allowing a deeper understanding of their design and operation.

5. Latency
Reduced Response Times: Running a model locally can minimize the delay caused by sending requests to a server and waiting for a response.
Real-Time Applications: This is especially valuable for applications that require real-time processing, such as voice assistants or robotics.

4. Offline Access
No Internet Dependency: A locally hosted AI can function without an internet connection, making it useful in remote locations or during outages.

3. Customizability
Fine-Tuning: Local models can often be fine-tuned or adjusted to meet specific needs, whereas hosted models are usually static and generalized.
Integration: You have full control over integrating the model into workflows, software, or hardware.

2. Cost Savings
No Subscription Fees: Once you've set up a local model, there are no recurring fees. This can be cheaper in the long run compared to subscription-based services.
Reduced Cloud Costs: For developers or businesses with high usage, local inference eliminates ongoing API or cloud costs.

1. Privacy and Data Security
Local Control: Running AI locally ensures your data doesn't leave your device, reducing concerns about data breaches or third-party access.

Running AI locally offers several advantages over using services like ChatGPT, Claude, or Gemini, depending on your needs, priorities, and constraints. Here are some key reasons:

just use a local llm

Are you using ChatGPT?

Pocket AI is like ChatGPT but it runs locally/offline on your phone to preserve your privacy.

Running Llama 3.2 3B locally on my iPhone 13 Pro at more than 30 tokens per second.