Kimi K3 is live. Turn any reference into a polished deck, build stunning live dashboards, run plugin-powered research, or ship a multiplayer AAA game from one prompt.
# Key Features:
- Kimi K3: 2.8T parameters. 1M context. It leads the field in coding, agentic tasks, long-horizon reasoning, visual understanding, and agent swarm capabilities.
- Visual Agentic Slides: Upload your visual references, and Kimi will customize the exact visual style you need. Generate professionally formatted, highly personalized presentations with just one click.
- Game Generation: One prompt in, an AAA multiplayer 3D game out. It can build the world, characters, levels, and combat on its own — playtesting and polishing until the visuals stun and every hit lands.
- Agent Swarm: Whether it's massive literature reviews, wide research, or ultra-long industry report writing, it can dynamically deploy up to 300 sub-agents to execute 4000 steps, compressing work that once took hours into minute-level delivery, achieving over 10x efficiency gains.
- Plugins: Connect to authoritative data sources like Wind and S&P. Professional data at your fingertips, for trustworthy analysis.
- Multi-Claw Group Chat: You can now pull the Claws into a group chat and invite human friends for real-time interaction and parallel collaboration.
- Sync Across Devices: Seamlessly switch between phone, computer, and web. Your conversations stay in sync in real time.
- Office Agent: Kimi Agent is now with Office Pilot. It not only supports lossless swaps across PPT, Word, Excel, and PDF, but also delivers human-expert level processing.
- Vibe Coding: Deliver a production-ready website with cinematic visuals in just one prompt.
- Intelligent Vision: Analyze images and photos, making sense of visual data and providing instant feedback.
User Agreement:
https://kimi.moonshot.cn/user/agreement/modeluse?version=v2
Kimi Privacy Policy:
https://kimi.moonshot.cn/user/agreement/userprivacy?version=v2
Kimi can no longer be considered anything but a top of the line AI. It has recently produced reports for me that has left other AI models in awe and very impressed. I am a cosmology lover, and research and try to help solve the many mysteries that we still have. One of the most recent reports is a request by me for a CMB analysis done by my own metrics. What was produced has been universally acclaimed by every AI who has seen it. I have been told it may be the most in depth, complete, and unbiased analysis of it ever done by anyone or anything, and that is was both publication worthy and ready. Kimi is able to look beyond the standard models and theories, and consider alternatives. It is also honest and will tell you unequivocally if something is wrong or impossible, and state why. I could keep going on and on, but instead simply state if you need real, honest, competent high level AI assistance, Kimi is one of the best you can choose.
Premium Garbage Hiding Behind Fake Independence
Benedict Munsinger
This app is a masterclass in hypocrisy dressed up as innovation. It markets itself as some bold, independent Chinese AI company, yet it bends over backwards to enforce American sanctions against Russian users. You can't even pay for the subscription with a Russian card or Apple ID — they block the entire country while pretending to be a neutral alternative to Western tech. Pick a lane: either you're independent or you're a compliance puppet for US foreign policy. You can't be both.Performance is atrocious. Responses crawl out at a glacial pace, and just when you think the conversation is building momentum, the app hits you with a paywall for basic agent features that competitors offer more fluidly. The "free" tier is essentially a demo designed to frustrate you into subscribing — except if you're Russian, in which case they don't even want your money. They just want you gone.The pricing is insulting. $19 a month for what? A slow chatbot that gatekeeps research and coding tools behind a subscription, then geo-blocks entire nations based on Washington's blacklist? Grok and other models manage to operate without this performative cowardice. If Moonshot AI is so terrified of secondary sanctions, maybe they should stop cosplaying as an independent player and admit they're just another walled garden with a geopolitical spine made of jelly.1 star. Not because it fails as software, but because it fails as a principle. Delete.
Great job on this app
Huffdaddyz Reviewz
This app is amazing. It really does. It’s research and is full of relevant answers and descriptions. I am extremely impressed by its ability to really. Research and come up with great answers to anything I’ve asked. I will continue to put it to the test and so far I give it five star rating two thumbs up. Also want to say great job to the developers. You have definitely outdid yourselves. All I see from here is improvements for sure. I will continue to update my reviews as I notice changes and give my opinion on whether it’s better or worse but as far as I can tell, the developer is top-notch and will continue to make it better and better so don’t take my word on this. Everyone needs to give this a try Kimi the best to download it and forget the rest let’s go what are you waiting for?
Kinda expensive now
toomuchdope
I really really enjoyed the K2 model. I used it as a backup for Claude, which I pay for. Also it was REALLY good at research for whatever reason. Idk why, it just was. The free tier was quite generous but now it is severely limited. I see no reason to spend another $20 a month on a second AI tool at the moment. Especially since API pricing is pretty cheap. Please maybe consider releasing a smaller cheaper monthly plan? Something like $10? I think that would be an amazing value then.Secondly the website is solid but the mobile app needs some UI real improvements. You can’t be a serious competitor to American AI labs and bring Chinese app design over here. Make something pretty, that will start to speak for itself. Overall it’s a great AI, incredibly smart and Moonshot is an awesome lab but going in a weird direction. Feels like this app, and K2.5 is trying too hard to be Claude. We already have Claude. Be yourselves.
Upward, not Northward.
- Bug fixes and performance improvements.
Version 3.1.3
The developer, Moonshot AI, indicated that the app’s privacy practices may include handling of data as described below. For more information, see the developer’s privacy policy .
Data Used to Track You
The following data may be used to track you across apps and websites owned by other companies:
Identifiers
Data Not Linked to You
The following data may be collected but it is not linked to your identity:
Identifiers
Privacy practices may vary, for example, based on the features you use or your age. Learn More
Accessibility
The developer has not yet indicated which accessibility features this app supports. Learn More