Ollama.com just started a 'transparency' campaign on token expenditure. Two months ago I left Claude for GLM+K+Deepseek clankers and have never once gotten close to blow the same budget on open-source paid inferences. Bare metal is the way, otherwise inference providers is the only comprimise. You can't measure quality/cost otherwise.

Think before reacting people! I'm a paraplegic, in a wheelchair. Shoelaces NEVER come undone, if you don't walk 🥴

"I changed the sidebar... six times" 😂

If privacy is not a concern to you, then you're winning. If it is, then you should audit the build with your llm about removing --features telemetry,nostr,otel prior to compiling it goose-cli.