Kimi-K3 narrows gap to closed-source AI frontier to 1.5 months
Open-source model now just weeks behind OpenAI and Anthropic in performance.
Deep Dive
According to the article, Kimi-K3’s release brings the open-source frontier to just 1.5 months behind closed-source models, per Artificial Analysis—putting it right on the heels of OpenAI and Anthropic. Despite this, the “it’s over for Anthropic” talk is considered overblown, as it’s still months behind the frontier. The model nears 3T parameters, proving scaling laws still hold, but its size makes local deployment impractical.
Key Points
- Kimi-K3 brings open-source within 1.5 months of closed-source frontier models.
- Model size approaches 3 trillion parameters, confirming scaling laws.
- Still behind Anthropic's Fable and OpenAI, with Google's Gemini 3 Pro absent since November.
Why It Matters
Open-source AI is catching up fast, threatening the lead of proprietary labs and democratizing access to cutting-edge capabilities.