Moonshot AI launches Kimi K3

Moonshot AI released its Kimi K3 model, a 2.8T parameter large language model the company said is on par with Anthropic's Claude Opus 4.8 and OpenAI GPT-5.5.

Kimi K3 lands as the open model battle kicks off in earnest. Even in the US, the land of proprietary, token budget busting models, has new open model players such as Thinking Machines, which just released its Inkling open-weight mixture-of-experts model that's trained for broad usage rather than optimized. The company outlined the details in a blog post.

SpaceXAI's move to open source Grok Build is another US open entry worth watching.

• Rightsizing open models may cut your AI inference spend
• SpaceXAI, Meta puts pricing squeeze on Anthropic, OpenAI
• Moonshot's Kimi K2.5 introduces agent swarm, highlights open source model momentum

Assuming Moonshot's benchmarking is correct, Kimi K3 may be shortening the distance between proprietary and open models. In a blog post, Kimi noted:

"While its overall performance still trails the most powerful proprietary models, Claude Fable 5 and GPT 5.6 Sol, Kimi K3 demonstrated frontier-level performance across our evaluation suite, consistently outperforming other tested models."

Kimi said that full model weights will be released by July 27.

Here's what you need to know about Kimi K3:

• Kimi K3 is a 2.8T-parameter model.
• There's a 1 million token context window.
• Vision capabilities are native.
• Kimi K3 is a scaled up mixture-of-experts model.
• Kimi K3 is built on recent architectural updates Kimi Delta Attention (KDA) and Attention Residuals (AttnRes).
• The model was built to optimize GPU kernels.
• Kimi K3 built a chip design based on itself in a proof of concept.

As for the pricing, which is the main thing being closely watched, Kimi K3 pricing is 30 cents per million cached input tokens, $3 per million non-cached input tokens and $15 per million tokens for output.

More:

• DeepSeek's real impact happening now
• Here’s what we learned about AI projects from enterprise buyers so far
• Where’s tokenomics for the rest of us?
• AI inference costs are going to be a big concern: What's the fix?
• Why enterprise AI leaders need to bank on open-source LLMs