After Chinese DeepSeek, which is to release anticipated DeepSeek V4 soon, we have Kimi K3 again from China!
Everyone in America must be thrilled, some anxious, some fearful, and some, like common people who pay for AI models, would be happy!
Kimi K3 was officially released by Moonshot AI.
2.8 trillion parameters.
World’s first open-source AI model.
The world needs open-source, cheaper models.
What a competition!
Complete model weights to be released by July 27, 2026.
Algorithm is their Kimi Delta Attention (KDA)
SDK and API provided for Kimi K3.
Works for text, image, and video input and outputs.
1 million tokens and caching.
The models are competitive with the best AI models as per the internet.
Would American companies lose their customers to China now?
How much profit can China make if it keeps releasing competitive AI models?
How can American companies retain their customers? Reduce rates per million tokens?
Would American AI companies become open source in the future?
What would we look for in future models from across the world?
We need models from around the Globe! Japan, Germany, or be it Brazil! This way, income can be distributed across the globe!
This is the next World Cup; let’s play who gets what! Else incomes would be disproportionately distributed! Let’s work hard now!
What would future AI models be like? I asked this of Kimi; here is a short reply.
- Multimodal.
Future models would combine text, image, audio, video, and even sensor data into a single reasoning process.
2. Long-Context as a Foundation, Not a Feature
Tomorrow’s models will treat “infinite” or near-infinite context as the baseline. They’ll be able to hold entire codebases, years of personal conversation history, or full-length research libraries in working memory, reasoning across them without losing coherence.
3. Agentic Autonomy
The shift from “chatbot” to “agent” is already underway.
4. Personalization Without Compromising Privacy
5. Faster, Cheaper, More Efficient
6. Reasoning as a Core Competency
7. Alignment and Safety
That was great Kimi !
Next, I asked Kimi, “Which country will earn the most from AI models in the future?”
The answer by Kimi is USA! Seems like Kimi doesn’t have confidence in itself? Why Kimi?
This chart is what Kimi gave me:
USA loses its proprietorship in the AI race! Europe, where are you?
Europe, France, AI model, Mistral AI, where are you?
Japan, South Korea? China is here now!
The big question is, would it be banned in some countries?
Would it raise cyber threats in some countries?
Is Kimi K3 safe to use?
Would Kimi K3 undergo Fable 5-like checks in countries for secure systems?
Thank you for reading
Subscribe for updates
Artificial Intelligence, Computer Science & Mathematics 