Moonshot AI has released Kimi K3, a new frontier model that approaches the performance of leading closed-source models. K3 boasts 2.8 trillion parameters, a one-million-token context window, and a Mixture-of-Experts architecture. While it trails top models like Claude Fable 5 and GPT-5.6 Sol in some benchmarks, it shows strong performance in agentic tasks and coding, and offers a significantly lower cost per token than its predecessors. The model is currently available via API and Kimi's products, with open weights expected by July 27, though demand has already strained its inference capacity. AI
IMPACT Sets a new benchmark for open-weight models, potentially accelerating competition and innovation in the LLM space.
RANK_REASON Frontier-lab model release with system card and benchmark comparisons. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
- Attention Residuals
- Claude Fable 5
- DeepSeek R1
- GLM-5.2
- GPT-5.6 Sol
- K2.6
- Kimi Delta Attention
- Kimi K3
- Moonshot AI
- Opus 4.8
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
