China's Moonshot AI launches open-weight Kimi K3 model closing gap with US rivals
New York, July 17
Chinese startup Moonshot AI launched its most capable frontier model, Kimi K3, which it claims can rival most of the advanced models by Anthropic and OpenAI.
The company said that the model is a 2.8 trillion-parameter model with a 1-million-token context window.
"It is the world's first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning," the company said in its blog.
The startup claimed that it demonstrated frontier-level performance across its evaluation suite but still trails Claude Fable 5 and GPT 5.6 SoI.
"Kimi K3 is available today on Kimi.com, Kimi Work, Kimi Code, and the Kimi API. At launch, Kimi K3 will use max thinking effort by default, with low- and high-effort modes to be introduced in subsequent updates," the company said. The full model weights will come out by July 27, 2026.
Arena.ai ranked the Kimi K3 model first in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5. This is a 17-place jump from Kimi-k2.6.
"In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5," Arena.ai said in a post on X.
Tech investor and US President Donald Trump's advisor David Sacks took to X to warn Americans that the Chinese were gaining an upper hand in the AI race, terming the development "concerning".
He said that AI development is being hampered by politicians and bureaucrats who are "banning new data centers, piling on state regulations and pushing for new federal agencies to pre-approve frontier models".
"This is how you lose the AI race. The rest of the world won't play by our rules if we bog ourselves down. Permissionless innovation is how America won the internet and became the technological envy of the world. We can do it again with AI -- while addressing risks in a targeted way -- or we'll watch our lead evaporate," Sacks wrote on X.
— ANI
Reader Comments
David Sacks is rightfully concerned. America is getting too caught up with regulations while China is just running ahead with open-source models. India should learn from this too - we need to encourage more research and fewer barriers. Jai Hind!
2.8 trillion parameters?! That's insane. I'm a software engineer and even I struggle to wrap my head around these numbers. The 1M token context window would be amazing for code analysis though. Let's hope this model is actually as good as they claim.
Honestly, I'm tired of all these AI companies hyping their models with benchmarks. Real-world performance is what matters. Indian startups should focus on building practical applications with these models rather than just chasing parameter counts. Kaam ki cheez chahiye!
The race is fascinating but I worry about the energy consumption. 3 trillion parameters means massive compute requirements. Is anyone tracking the environmental cost? Maybe we need to balance innovation with sustainability.
Good for China but India needs to step up its game fast. We have the talent but our policies are too slow. Look at how quickly Moonshot AI went from research to deployment - that's the speed we need. Better late than never, but let's not be too late.
We welcome thoughtful discussions from our readers. Please keep comments respectful and on-topic.