Moonshot's Kimi K3.1 Pro claims a lead over Claude and GPT, and China's open frontier keeps closing in

Chinese company Moonshot AI has launched Kimi K3.1 Pro. According to the information available, the model has 2.8 trillion parameters, the internal weights that set its size. Moonshot says it outperforms Claude Opus 5.5 and GPT-6 Astra on a range of performance tests. Our source is only a brief note. It does not say which benchmarks were used or by how much the model wins.

**Our reading.** The launch fits the trend we have been tracking: Chinese labs keep closing the gap with the Western frontier, step by step. Kimi K2.6 already sat close behind on our own rankings. But "beats X on several tests" is the framing we have learned to question. What matters is which benchmarks were chosen, and whether the results hold up on hard tests run by independent evaluators rather than ones picked by the developer. Until those results arrive, treat this as a claim, not a proven capability. Either way, Chinese pressure on the frontier is good news for competition and for lower costs.