DeepSeek V4 Pro trails Kimi K2.6 in benchmark, shows gains
DeepSeek's flagship V4 model has underperformed against domestic and US rivals, according to a new analysis by Artificial Analysis. The company's most advanced system, V4 Pro, ranked second among open-source models globally, behind Beijing-based Moonshot AI's Kimi K2.6. While V4 Pro showed clear improvement over its predecessor V3.2, it still lagged top competitors. The results highlight challenges facing DeepSeek as China races to close the AI gap with the US amid intensifying competition and computing power constraints. Analysts noted that V4 delivered meaningful technical progress, raising hopes for another breakthrough similar to the earlier R1 release.
Key facts
- DeepSeek V4 Pro ranked second among open-source models globally.
- V4 Pro trailed Beijing-based Moonshot AI's Kimi K2.6.
- V4 Pro showed improvement over predecessor V3.2.
- Analysis was conducted by benchmark firm Artificial Analysis.
- Report was released on Friday.
- DeepSeek faces challenges from domestic and US competition.
- Computing power constraints affect DeepSeek's progress.
- Analysts noted meaningful technical progress in V4.
Entities
Institutions
- DeepSeek
- Moonshot AI
- Artificial Analysis
Locations
- Beijing
- China
- United States