I use the venice.ai for private AI mostly (currently using GLM 4.7 and qwen3 next). They even got anonymized claude, GPT and others if you really want that.
I got some of their crypto to get free API credits each day, so I'm running it essentially "free" because I can get that money back if I want.
I chat with the AI via my selfhosted open-webui instance, installed as PWA on my phone.
If you don't want to invest in that ecosystem, you can also just use the free or pro version normal app or PWA. But not so much subscription based version of their API.
Oh and btw, you can privately use deepseek v3.2 there if that's something you care. You meantioned it in your question so