Unverified50% confidenceBenchmarkExact time
After Kimi K3 was released in 2025, it ranked among the top open-source models on multiple benchmarks, with particularly outstanding performance on mathematics and code reasoning tasks.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
9/24/2026
First Seen
Valid until: 12/23/2026
Sources
Related Entities
Related Claims
UnverifiedKimi K3 is a 2.8T-parameter multimodal LatentMoE model that uses 896 experts with top-16 routing and performs routing computation in a 3584-dimensional latent space.68% similarUnverifiedKimi K3于7月16日发布并接入Kimi Code，采用2.8万亿参数规模，最高支持100万Token上下文和原生视觉理解67% similarVerifiedKimi K2.6 adopts a Mixture of Experts (MoE) architecture, with a total parameter count at the trillion level and only 32B parameters actually activated.67% similarUnverified以Kimi K3实验，给它预填充1%的Claude Opus解码推理片段，其输出风格就会明显向Opus靠拢，某些区段Claude推理链从Kimi K3提取的容易程度比其他模型高出约6个数量级67% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/946748API
curl https://kongchang.com/api/v1/knowledge/claims/946748MCP
get_claim(id=946748)