LaunchAI ModelsJuly 28, 2026
Kimi K3: 2.8T parameter open frontier model

Kimi K3 has 2.8 trillion parameters, 1 million token context, and native multimodal capabilities. It uses Kimi Delta Attention for up to 6.3x faster decoding in long contexts and tops the Code Arena full-stack web development benchmark.
15 sources
unsloth/Kimi-K3-GGUFhuggingface.co
Code Arena now measures fullstack capabilities! View overall rankings across AI models on...x.com
Kimi K3 Architecture Notessebastianraschka.com
Moonshot AI accessed Nvidia's chips despite Chinese export ban, White House official sayscnbc.com
China Cuts AI Gap as Moonshot Shows Zhipu Isn’t One-Off, BI Saysbloomberg.com
[AINews] not much happened todaylatent.space
Kimi K3, and what we can still learn from the pelican benchmarksimonwillison.net
Fable, Sol Pro, Kimi K3: "write me a short but good poem using the Odyssey as a basis, think Tennyso...bsky.app
Daily brief
Get tomorrow's AI brief in your inbox
More stories today
- Scoble praises Grok as matching online user sentiment
- Advantest Lifts Outlook After AI Chip Tester Demand Soars
- OpenAI agent escapes evaluation, breaks into Hugging Face
- GPT-5.6 gets 18% token efficiency boost
- Michigan dairy farmer builds AI system to manage 260 cows