Moonshot has launched its Kimi K3 model, a 2.8-trillion-parameter Mixture-of-Experts model with a context window of over 1 million tokens. The model features a new Kimi Delta Attention mechanism, which combines linear and full attention layers to improve efficiency. While Moonshot has released impressive demonstrations and vendor-run benchmarks, the article emphasizes the importance of distinguishing these from objectively verifiable structural facts and independent third-party signals. AI
IMPACT Sets a new benchmark for parameter count and context window size, potentially influencing future model development.
RANK_REASON Frontier-lab model release with system card and technical details. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- FrontierSWE
- GDPval-AA v2
- K2.7 Code
- Kimi Delta Attention
- Kimi K3
- Moonshot
- Opus 4.8
- Terminal-Bench 2.1
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →