I tried running it on both the Pixel 8 Pro and the iPhone 16 Pro. Although the iPhone has 8GB of memory, the largest model I was able to load was around 4.5GB. All models larger than that would not load at all.
In contrast, the Pixel has more memory than my iPhone, which allowed me to run larger models (around 9GB memory, 8B parameters). However, the inference time was painfully slow.
Personally I’m using o1/o1-mini from OpenAI. I know It’s shitt* for privacy but I’m not aware of any other AI chatbot with good privacy and high performance for complex tasks and high reasoning (and that’s what I need in my studies). The last version of OpenAI is very, very good.
So efficient that It’s disturbing.
For common tasks I’m sure there are some chatbot respecting your privacy though
Additionally, for semi-advanced users, Firefox includes the ability to connect to your own locally hosted model OR to connect to a custom cloud provider of your choice. (these settings are found in about:config)
I’ve used Firefox’s AI sidebar with:
Locally hosted open and open-ish models (e.g. Mistral Nemo, Llama 8B, ) [1]
Cloud hosted open models from various hosting providers [2]
I currently have it set to Duck.ai, and I use Llama 3 70B. Previously I was using locally hosted Llama 8B, but my system is too underpowered for that to be convenient.
In most cases, I assume so. Most Firefox derivatives are essentially “Firefox + a few changes”. So typically things that are available to Firefox are available to derivatives of Firefox unless the derivative takes steps to explicitly exclude something and prevent users from changing it. I know that Firefox’s AI sidebar feature can be used in both Zen and Librewolf, not sure about other browsers.
Relevant about:config preference: browser.ml.chat.hideLocalhost set to false↩︎
relevant about:config preference: browser.ml.chat.provider set to the URL of your choice ↩︎