Kimi K3.1 Pro hands-on: roughly on par with Gemini 4 Argon
A tester says Kimi K3.1 Pro is a big quality jump with fewer hallucinations and lands around Gemini 4 Argon, with each model winning different tasks.
okay, first thoughts on Kimi K3.1 Pro after testing it. it's actually pretty damn good ngl. massive improvement over the older generation, hallucinates a lot less too. I'm getting around 40-50 tokens/sec, which isn't particularly fast, but in my tests it seems to use fewer tokens to finish tasks now, so the overall completion time is better. quality wise, it's way ahead of DeepSeek V4.1 Flash 1017, although quite a bit more expensive. it straight up mogs GPT 6.1 Sol 😭 I'd put it somewhere around Gemini 4 Argon. Some tasks Argon does better, others Kimi wins. genuinely can't decide between the two yet. my current picks: → Opus 5.5 is still KING if you're using subscriptions. → DeepSeek V4.1 Flash 1017 is KING for cheap API usage. → Kimi K3.1 Pro is my new KING if you want high quality through an API without paying for a subscription. and now we wait for Fable 5.5, hopefully in the next few days. Anthropic might just take that crown back again 😭
This post quotes another post: x.com/sahilpanhotra/status/2113932250622611733