LaunchAI ModelsJuly 24, 2026
Moonshot AI launches Kimi K3, a 2.8T parameter open-weight model

Kimi K3 features 2.8 trillion parameters, 1 million token context, and native multimodal capabilities with Delta Attention for 6.3x faster decoding. In DeepSWE benchmarks, it performs near Claude Fable 5 at ~35% of the cost, with a 2.8x higher solve rate per dollar.
15 sources
Kimi K3 vs Claude Fable 5 on DeepSWE: Cost and Codingtogether.ai
We ran GLM 5.2 and Kimi K3 on 100 deep research tasks (DRACO by perplexity) and asked Fable to be...x.com
China Cuts AI Gap as Moonshot Shows Zhipu Isn’t One-Off, BI Saysbloomberg.com
Moonshot is Chinese But Its AI Models Are From Another Planetthealgorithmicbridge.com
[AINews] not much happened todaylatent.space
Chinese startup Moonshot AI unveils Kimi model it says rivals OpenAI, Anthropiccnbc.com
Kimi K3, and what we can still learn from the pelican benchmarksimonwillison.net
More stories today
- Vercel co-signs Open Weights and American AI Leadership letter
- How Autonomous AI Is Transforming Chip and System Design
- NVIDIA expands Agent Toolkit with PhysicsNeMo and CUDA-X libraries
- NVIDIA advances semiconductor materials engineering and manufacturing
- NVIDIA Nemotron 3 Ultra achieves benchmark-leading performance with LangChain Deep Agents