That's no surprise. Just install Ollama and it will run on almost anything, as long as the LLM fits in memory.
Stable Diffusion XL Turbo on a 512MB Pi Zero 2, now that's an achievement (but you need a lot of patience)! https://github.com/vitoplantamura/OnnxStream
As far as I can tell 'no' or not to the same extent. Typically (GPT / CoPilot) censoring is done as a first-pass on a special detection system before hitting the LLM, probably similar with the app/site here. I used prompts about Tiananmen and Taiwan and they gave *tilted* answers.
What I mean by that, is that they were definitely biased pro-gov, but it wasn't direct, it was subtle. The output was similar to large-scale LLMs, but would do things like ... present the topic of Taiwan to appear neutral, but use most of the text space to describe China's position.
It does also refuse to consider (or minimizes) new information I give it that contradicts things it thinks it knows. For example feeding it some of the content of Trump's executive orders, it minimizes or re-interprets the information based on what it already "knows" - which here is very different
i asked it how many 'r's are in strawberry and it spat out 4 pages of thinking, in which it somehow gaslit itself into the wrong answer
this is still just a spicy autocorrect plagiarism machine, now with extra chinese censorship
My local DeepSeek-R1-Distill-Qwen-7B instance says:
Based on the analysis of "Hailey" in relation to Kiki and Bouba from "The Wonder Years," it is clear that "Hailey" does not align with either character.
Answer: "Hailey" refers to no specific character in "The Wonder Years."
Comments
but I might be getting confused with Sam’s alt which was a chatbot for a bit lol
But otherwise, from what I've seen on my desktop - yes. I'm getting GPT-class performance on a 12GB video card, no RAM.
... w/some interesting typos in Chinese.
that doesn't surprise me 😂
Runs slower when I give it 8k or 16k context but still usable
Stable Diffusion XL Turbo on a 512MB Pi Zero 2, now that's an achievement (but you need a lot of patience)!
https://github.com/vitoplantamura/OnnxStream
this is still just a spicy autocorrect plagiarism machine, now with extra chinese censorship
Based on the analysis of "Hailey" in relation to Kiki and Bouba from "The Wonder Years," it is clear that "Hailey" does not align with either character.
Answer: "Hailey" refers to no specific character in "The Wonder Years."
or the easier way, https://github.com/skyware-js/labeler