Comment by iLoveOncall
12 hours ago
I might be wrong but subscription models don't give you API access. I'm only interested in API access when it comes to personal use, so local models running for free makes sense for me.
That said I have an RTX 5090, not a Mac Mini, so it's not exactly the same level of performance... The latest open models run at 200 tpm at around 30B params.
Only Anthropic does that AFAIK, at least I can use my $20/month Codex and Kimi subscriptions in pi.