Not my experience, I've found it extremely useful for when I need a quick answer. It's handled very specific fairly obscure questions well. I also find it quite consistent.

▲

HDBaseT 8 hours ago | parent [-]

It hallucinates about 50% of the time. [0].

It is completely useless and worse than quantized gpt-oss from over a year ago. It is designed for large scale, instant responses.

I hate when co-wokers or friends say "when I searched it up" and proceed to spew out whatever the Google AI said. I would consider listen if they ran it through ChatGPT, Gemini or Claude but whatever model is used on the search results is just bullshit.

[0] https://oumi.ai/blog/oumis-study-finds-50-of-ai-overviews

	▲	x______________ 3 hours ago \| parent [-]
		> It is designed for large scale, instant responses. Who wins when someone has to make 3 or 4 prompts to get an answer as opposed to just the initial prompt? Seems likr studies like this need to be made at regular intervals to keep tabs on model functional, especially public facing ones..