China’s new heavyweight challenger
Moonshot AI just dropped Kimi K3, a 2.8-trillion-parameter model that’s squarely aimed at the top of the AI food chain. It’s built for coding, research, reasoning, and vision tasks, and the company isn’t shy about comparing it to the best closed models from OpenAI and Anthropic.
The catch? Moonshot itself admits K3 still trails Claude Fable 5 and GPT 5.6 Sol overall. But the benchmark numbers tell a more interesting story.
How close is Kimi K3 to the best closed models?
On several key tests, K3 actually edges out both rivals. It scored 77.8 on Program Bench, narrowly beating Fable 5’s 76.8 and GPT 5.6 Sol’s 77.6. It also led BrowseComp with 91.2 and SWE Marathon with 42.0, ahead of both.
Other results show the gap hasn’t closed entirely. K3 scored 67.5 on DeepSWE, versus 70.0 for Fable 5 and 73.0 for GPT 5.6 Sol. Moonshot also says the overall user experience still lags behind the proprietary pair.
There’s an important caveat to all this. Moonshot notes that Fable 5 results may include fallbacks to another model, and GPT 5.6 Sol results may include cyberguards that restrict certain responses. In other words, the comparison is useful — but not perfectly even.
Open weights put pressure on OpenAI and Anthropic
The most significant part of K3 might not be its performance at all. Moonshot plans to release the full model weights by July 27, meaning you can download and run K3 locally — provided you have the hardware to handle a model this size. Developers can also modify and fine-tune it for specific tasks.
That open-weight approach is a direct challenge to the closed strategies of OpenAI and Anthropic. When a near-frontier model is freely available, it gets harder to justify the premium prices those companies charge.
Pricing that undercuts the U.S. giants
K3 isn’t the cheapest Chinese model on the market, but it’s still significantly more affordable than its U.S. counterparts. Its API pricing is $0.30 per million cached input tokens, $3 for uncached input, and $15 for output. That puts it closer to Anthropic’s mid-range offerings than to the rock-bottom prices of earlier Chinese releases.
Chinese AI models have long been cheaper than American ones, and that gap is already pushing some U.S. startups to adopt them to cut costs. Kimi K3’s near-frontier performance at a lower price only adds to that pressure.
What this means for the AI race
The release of Kimi K3 signals that the competitive landscape is shifting. Open-weight models are no longer a step behind on every metric — they’re trading blows with the best closed systems on specific benchmarks. For developers who want control and cost savings, that’s a compelling combination.
It also raises a question: how long can OpenAI and Anthropic maintain their premium pricing when a model like K3 is close enough on performance and free to download? The answer may determine the next phase of the AI market.
If you’re weighing your options, it’s worth keeping an eye on how K3 performs in real-world use, not just on benchmarks. And for those who prefer the closed-model route, the pressure from open-weight rivals could lead to better pricing or features down the line.
