hotAI

2 min read

Moonshot’s Kimi K3 puts pressure on OpenAI

Moonshot says its new Kimi K3 model rivals top US systems, and its open-weight release could make it especially disruptive.

Image: TechRadar

Moonshot has unveiled Kimi K3, a new model the Chinese company says can compete with leading US systems such as ChatGPT and Claude. According to Bloomberg, Moonshot’s own benchmarks show Kimi K3 trailing only Claude Fable 5 from Anthropic and GPT-5.6 from OpenAI.

Bloomberg says Kimi K3 has 2.8 trillion parameters, while Artificial Analysis ranked it ahead of Anthropic’s Opus 4.8 on some benchmarks. Moonshot also says the model beats Chinese rival Z.AI on coding tasks. The company describes it in a blog post as the “world’s first open 3T-class model, designed for frontier intelligence across long-horizon coding, knowledge work, and reasoning.”

The model is available now, and Leonid Mironov, a portfolio manager at Gavekal Capital, told Bloomberg:

Recommended reading

Apple broadens OpenAI fight as hardware ambitions grow

“In my use, it’s clearly the best Chinese model ever.”

Leonid Mironov, portfolio manager at Gavekal Capital

Why the open-weight release matters

The letters AI in a box in the middle of a vast digital room divided by beams of line
The letters AI in a box in the middle of a vast digital room divided by beams of line

What makes Kimi K3 especially notable is that it is open weight. From July 27, anyone can download the model weights and run it themselves without paying for the model. That does not make it open source: users get the trained model, but not the underlying training methods or data.

There is a practical limit. Running a model of this size requires extremely powerful hardware, so the no-cost access is most meaningful for companies with deep infrastructure budgets. Even so, an open-weight model that performs near the top of the market is a direct challenge to US leaders such as OpenAI and Anthropic.

Moonshot is also targeting one of the most commercially attractive parts of the market: coding. The report notes that Moonshot charges more than Chinese rivals for access, with pricing said to be around Claude Sonnet levels, putting it alongside current frontier models rather than discount alternatives.

That combination — high-end performance, coding focus, and open-weight distribution — helps explain why Kimi K3 has drawn attention. It also arrives amid broader concerns that intensifying competition could push safety protections aside in the race for faster progress. Bloomberg further notes that the US State Department earlier this year accused Chinese firms, including Moonshot, of AI theft.

Ava Chen

AI Editor

Ava covers the rapidly evolving world of artificial intelligence, from foundational models and research labs to the real-world economics of intelligence. With a background in computational linguistics, she cuts through the hype to find out what actually works. She firmly believes that benchmarks are just marketing until reproduced in the wild.

via TechRadar

// Keep reading