China’s AI model arrives to challenge US:Kimi K3 has human-like eyes and brain; will do coding and research in minutes

Chinese company ‘Moonshot AI’ has introduced its new AI model Kimi K3. The company claims that it is the world’s first open model with 2.8 trillion parameters. It has a context window of 1 million tokens. Its scaling efficiency is about 2.5 times better compared to the previous model Kimi K2. Let’s understand this with a simple example: imagine an AI model is like a human brain that has to prepare for a very big exam. 1. What are Parameters? You can consider parameters as the ‘experience’ of AI. When a child grows up, the more things they learn, the sharper their mind becomes. In the world of AI, a model with more parameters is considered more intelligent. Having 2.8 trillion parameters means that this model has an extremely vast mind and can understand and solve very complex and difficult questions like humans. 2. What is 3T Class? Here ‘T’ means Trillion. Since this model has 2.8 trillion parameters, which is very close to 3 trillion (3.0T), it has been called a ‘3T class’ model. Reaching the 3T class in the world of AI is a huge achievement, as such large models were previously only available to select companies like America’s OpenAI. 3. What is a 1 Million Token Context Window? Context window means AI’s ‘memory’ or its ability to read at once. When you chat with AI, how long a paragraph or document it can remember at once, that is its context window. Example: If you give a common AI model 5-10 pages, it will read them, but if you give it an entire book, it will be overwhelmed. Kimi K3’s 1 million token capacity means that you can give it an entire thick book, thousands of lines of computer code, or a year’s worth of bank statements all at once; it will process all that information in its mind in a second and provide you with an accurate analysis. 4. What does 2.5 times better scaling efficiency mean? ‘2.5 times better scaling efficiency’ means that Moonshot company has made its software so efficient that it can perform larger tasks much faster, with two and a half times less computer power and cost compared to its previous model. 5. What is the advantage of being an ‘open model’? Until now, American companies have kept their large models hidden. This is called a closed model. They only let you use it, but don’t tell you how it was built internally. A Chinese company has kept it an ‘open model’. This means that it will freely or openly release the ‘blueprint’ of this powerful mind to software developers and scientists worldwide. This allows any ordinary engineer in the world to modify this AI according to their needs, create new apps, or use it in their work. Moonshot Will Open Its Secret Formula on July 27 Moonshot AI stated in its blog that Kimi K3 is built on ‘Kimi Delta Attention’ and ‘Attention Residuals’. It also has native vision capabilities. This model is designed for complex tasks such as coding, knowledge work, and reasoning. According to the company, its full model weights will be released by July 27. 1. What are ‘Kimi Delta Attention’ and ‘Attention Residuals’? Understand with an example: When you read a very thick book, do you pay equal attention to every single word? No. Your brain only ‘attends’ to the main words and important lines, and skips unnecessary words. Benefit: The new formula developed by Moonshot AI, named ‘Kimi Delta Attention’ and ‘Attention Residuals’, gives this AI precisely this superpower. When you give it a very large task or a file with millions of words, this technology prevents the AI’s mind from wandering. It focuses only on the most important points and retains the ‘effect’ (residuals) of previously read information in its mind. This allows the AI to work very quickly without getting tired and without making mistakes. 2. What are ‘Native Vision Capabilities’? Understand with an example: Earlier AIs could only understand text. If you showed them a photo, they first had to convert that photo into words before they could understand it. But ‘Native Vision’ means that Kimi K3 has the ability to see. Benefit: If you show it a photo, a complex graph, a medical scan (like an X-ray or MRI), or a video, it will instantly understand it like human eyes and answer your questions about it. 3. Complex Tasks: What do Coding, Knowledge Work, and Reasoning mean? The company states that this model is built for the highest level of AI, which will perform these three major tasks: Long-Horizon Coding: This doesn’t just mean writing small bits of code. Like software engineers, it can write entire large computer programs or apps consisting of thousands of lines at once, and find and fix flaws in them. Knowledge Work: It can work like researchers, scientists, and senior lawyers. Sifting through years-old files, analyzing major legal cases, or preparing a complete research report on a subject is a breeze for it. Reasoning: It means ‘the ability to think, understand, and reason’. It will not just state memorized facts, but like humans, it will help in making the right decisions by considering ‘ifs and buts’ in difficult math problems or business complexities. Kimi K3 Lags Behind America’s Latest AI Models The company admitted that Kimi K3’s performance lags behind models like America’s Claude Fable 5 and GPT 5.6 Sol. However, in several benchmark tests, it has surpassed older American models. Moonshot AI is backed by Chinese tech giants Alibaba and Tencent. Chinese Firms Stole Data by Creating 24,000 Fake Accounts Meanwhile, American companies have expressed concern over Chinese firms releasing models at low prices. Anthropic has alleged that DeepSeek, Moonshot, and MiniMax launched large-scale campaigns to illegally extract the capabilities of their models. According to Anthropic, Claude was used 16 million times from approximately 24,000 fake accounts.