Moonshot AI Releases Kimi K3
Moonshot AI released Kimi K3 today, a 2.8 trillion parameter model that ranks 4th on the Artificial Analysis Intelligence Index, just a couple points behind Claude Fable 5 and GPT-5.6 Sol.
Benchmarks
Kimi K3 scores 57 on the Intelligence Index, placing it 4th overall. The top three are Claude Fable 5 (59.86), GPT-5.6 Sol (58.89), and Sol xhigh (57.65) — the gap is minimal.
It ranks 1st on the Front End Code Arena with 1679 points, surpassing Claude Fable 5. This is the first time a Chinese model has topped the coding leaderboard.
The 1M token context window continues Kimi's tradition of long-context capabilities.
Specs and Pricing
- Parameters: 2.8T (proprietary, closed source)
- Input: text + image
- Output: text
- Pricing: $3/1M input tokens, $15/1M output tokens
- Speed: 62 tokens/s (below average)
- Time to first token: 1.99s (better than average at 2.62s)
Pricing is above the market median for both input ($1.75 median) and output ($8.40 median), but reasonable given the 2.8T parameter count.
What It Means
Kimi K3 is the first Chinese model to break into the global top 3 intelligence rankings. The top spots have long been held by Anthropic, OpenAI, and Google. This release shows Chinese models can compete at the frontier in both base capability and coding performance.
On the Artificial Analysis leaderboard, Kimi K3 (57) trails GPT-5.6 Sol (58.89) by less than 2 points — potentially catchable in the next iteration.
Kimi K3 is available now through the Kimi website and API.