The Chinese AI model Kimi K3.1 Pro from Moonshot AI has made a significant jump in the Artificial Analysis Intelligence Index. According to The Decoder, it scores 55 points and ranks 4th. That is 11 points more than its predecessor.
Who is still ahead of Kimi
According to the report, only three models from Anthropic and Meta are still ahead of Kimi K3.1 Pro. The article's title names Anthropic's Claude Opus 5.5 as the benchmark that the Chinese model almost catches up to. An index like this condenses a model's results into a single number. That makes comparisons easy to read. Still, it should be seen as a snapshot. How well a model fits in everyday use depends heavily on the specific task.
Open weights as a second point
What Moonshot has planned next is also interesting. If the company releases the weights as planned, the gap between open and proprietary models would shrink to just three points.
"Weights" refers to a model's trained parameters. If they are public, others can in principle run, inspect and adapt the model themselves. With proprietary models, that is not possible. They can only be used through the provider's interface.
What readers can take away
For developers and companies, an open model close to the top can be interesting. They would get more control over where and how the model runs. Whether that applies here, however, depends on whether the release actually happens and under what conditions. The source gives no details on this.
The report is also very short. Anyone who wants to use Kimi K3.1 Pro should therefore test it on their own tasks and not rely on the ranking alone.
