Kimi K2 is so good at tool calling and agentic loops, can call multiple tools in parallel and reliably, and knows "when to stop", which is another important property. It's the first model I feel comfortable using in production since Claude 3.5 Sonnet.

been testing kimi k2 for mario (my sales agent) and honestly? the tool calling is chef's kissswitched from claude for one specific workflow where we needed 3 parallel api calls conditional logic. kimi just... handles it. no weird loops, knows when to stopstill claude for most things but kimi's nailing the complex agentic stuff

you don’t feel comfortable using Opus 4 in prod? 🤯 hype!!!

More so the first non-Anthropic model

Agreed - given it a proper go today and am astounded by the quality. General vibe checks have it providing better answers than o3 pro and Grok 4 on a bunch of tasks

When you say the last model you were comfortable using in production was Claude 3.5 sonnet, does that mean you did not use Claude 4 for production?

I didn’t word it properly, I meant it's the first non-Anthropic model I feel I can use for agentic loops.

The "call multiple tools in parallel" that is done on the server side of Kimi, not your own implementation, right? This means that when 3rd parties host Kimi K2, we will not have that tool calling capability?

You just build that

what are you using to run it? is it local?

No @OpenRouterAI. You can't really run this locally unless you have insane specs.