Moonshot AI unveils 2.8T MoE Kimi K3
Chinese AI startup Moonshot AI has released Kimi K3, a massive 2.8-trillion-parameter Mixture of Experts (MoE) open-weight model featuring a 1-million-token context window. The release represents a major advancement in open-weight models, showcasing frontier-level capabilities and intensifying the compute race between U.S. and Chinese AI labs.
The launch of Kimi K3 signals that Chinese labs are rapidly closing the gap with U.S. proprietary models by leveraging massive MoE architectures, shifting the competitive landscape toward high-performance open-weight models.
- –The 2.8-trillion-parameter MoE architecture showcases significant engineering and compute capability.
- –Offering a 1-million-token context window in an open-weight model lowers the barrier for developers building complex, long-context applications.
- –U.S. labs face mounting pressure to scale and release next-generation models as international competitors match frontier capabilities.
DISCOVERED
1h ago
2026-07-17
PUBLISHED
1h ago
2026-07-17
RELEVANCE
AUTHOR
pequityresearch