Moonshot AI, a company based in Beijing, is preparing to release a new AI model called Kimi K3. The model is very large, with between 2 trillion and 3 trillion parameters. For comparison, parameters are the internal settings a model learns during training, similar to how a brain forms connections. More parameters generally mean a more capable model. Kimi K3 is expected in the coming days as of July 16, 2026.

People familiar with the matter told the Financial Times that the model is expected to perform as well as or better than Anthropic's Opus 4.8, which is one of the most advanced AI models currently available. TechCrunch also reported the story.

Kimi K3 will be the largest open-weight AI model from China. The Financial Times describes it as China's largest AI model to date.

The open-weight label is important. It means the model's trained parameters will be available for anyone to download. Most top-tier AI models today are only available through a paid online service, where you send requests and get answers back. An open-weight model is different. Think of it like the difference between buying a car you can open the hood of and modify, versus taking a taxi where you ride but never see the engine. With an open-weight model, developers can change it, adapt it, and run it on their own machines.

There is a catch, though. Running a model this large requires serious computing power. Even a company with substantial resources would need a large number of specialized computer chips to operate a 3-trillion-parameter model in everyday use.

The expectation that Kimi K3 will match or exceed Opus 4.8 sets a specific performance target. Opus 4.8 is Anthropic's current top model, and closing that gap would put a Chinese open-weight model in direct competition with the best proprietary systems. Whether Kimi K3 actually delivers on this expectation will only be confirmable once independent testers evaluate it after release.

At the same time, Moonshot AI is reportedly trying to raise new funding at a $31.5 billion valuation, according to TechCrunch. That is a big jump from May 2026, when the company raised $2 billion at a $20 billion valuation. Going from $20 billion to a reported $31.5 billion target in about two months suggests investors are closely following the company's progress. Whether the new funding round actually closes at that level is not yet confirmed, since the figure comes from anonymous sources.

The Financial Times originally reported the Kimi K3 story on July 16, 2026. Both the Financial Times and TechCrunch describe the release as imminent rather than officially announced. Moonshot AI has not, according to the available reporting, issued a public statement with a specific release date.

The broader picture is what a model like this could mean for the AI world. If Kimi K3 performs as well as sources expect, it would narrow the gap between downloadable, modifiable models and the proprietary systems locked behind paid services. The open-weight community has been working with capable but smaller models. A release of this size with strong performance would change what people think is possible for open AI.

There is a geopolitical angle to consider as well. Chinese AI labs have been releasing more open-weight models lately, and Kimi K3 would push that trend to a new scale. For companies that face restrictions on using U.S.-based AI services, or that need to keep their data within certain countries, a downloadable model of this quality offers a way to get top-tier AI without relying on American providers. That is a real, practical option for organizations weighing their choices, not a hypothetical one.

The funding and the model release are, in all likelihood, connected. A model that matches Opus 4.8 and is freely downloadable would give investors a concrete reason to value the company at $31.5 billion rather than $20 billion. Investors betting on that price are counting on Kimi K3 performing as well as the sources say it will.

The coming days will tell whether the reported expectations become a released model, and after that, whether independent testing backs up the performance claims. For now, what is known is the company's intent, the model's size, its open-weight availability, a specific competitive target, and an aggressive fundraising context. The rest awaits the release.