At this point, running Chinese model like GLM-5 or Kimi K2 would be far more safer than risking off your LLM subscriptions. Quite the irony that our AI techno-feudal corpo overlords doesn't want to see their LLM take off with curious and useful open source ideas. Just like Microsoft, they deliberately buried it for some reason.

Oh, maybe not, they did it in the name of "terms of service abuse" and "risk assessment".

Thus it would be far better if we can just have SOTA open weight model to run OpenClaw/Clawdbot/Molt at least we are under control. And as you see the two Chinese models I mentioned are indeed open weight, albeit taking atrocious amount of resource to really self host, and you probably need to have abliterations to remove their political guardrails.

Sigh. We can't have great things with those big tech corpos and CCP politics. Big question: Why has this world gone to shit lately.