What model is the Research Assistant based on?

rogue

https://help.kagi.com/kagi/ai/kagi-research.html
Why doesn’t this list the base models that research assistant uses ?

Is the base LLM from OpenAI? Anthropic? X?
What does the image generation?

This is particularly of note because in the “LLM’s are bullshitters” article, the author notes that you should ask “Who is your LLM working for?”

Additionally, in the Assistant view, I think that all the models from Kagi should tell me what model(s) were used to generate the response.

It used to tell you the model used when you clicked the (i) icon, but now it doesn’t.
I know the models are tweaked by the Kagi team, but they’re not built by Kagi from the ground up, so they’re based on something and have all the inherent biases the creators input.
I would also like to know who the information I put in is going to. I know it’s anonymised , but it still is going to that provider.

Pleonasm

...And, what AI model(s) is used by the Research (Experimental) Assistant?

Since the answer may vary over time, there should be a straightforward mechanism in the Assistant interface to retrieve this information whenever it is of interest to the customer.

Rehan

I think you're right in that this should be more transparent to users. Having it in the documentation adds a bit of friction, as we risk having the documentation out of date as things change. I think ideally it should maybe live somewhere in the frontend. However, in the mean time, I'll implement it as a slash command. So you could type something like /info while using an assistant, and get back what models are used.

Pleonasm

@Rehan, thank you, in advance, for addressing the issue of allowing customers to view the model used by the Assistant and the Research Assistant. When the "/info" command has been enhanced to display this information, please reply to this thread to inform the community.