Kimi 3 Closing in on Opus 4.8: Reshaping Valuation Logic During China's AI Catch-Up Phase

According to the latest data from third-party benchmark platform OpenBench, Moonshot AI's Kimi 2 scored 87.3 on MMLU (Massive Multitask Language Understanding), while Anthropic's Opus 4.8 hit 92.1—a gap of about 4.8 percentage points. However, in HumanEval code generation tasks, Kimi 2 has caught up to 89.6 vs 91.2, narrowing the gap to just 1.6 percentage points. TechCrunch reports that the upcoming Kimi 3 is expected to match or even surpass Opus 4.8 on multiple core metrics, meaning the generational gap between Chinese AI labs and the global top tier has compressed from model generations (~1.5 years) down to version levels (~3 months).