Moonshot Releases Kimi K3, the Largest Open Model Yet at 2.8 Trillion Parameters

Chinese lab Moonshot AI unveiled Kimi K3 on July 16, 2026, a language model with 2.8 trillion parameters that the company calls the first "open model in the 3-trillion class." Moonshot promises to publish the open weights by July 27, 2026. For now, the model runs only through the company's website and API.

What we know about Kimi K3

Kimi K3 is Moonshot AI's largest model to date: 2.8 trillion parameters, more than double the previous trillion-parameter Kimi. By size, it outstrips DeepSeek V4 Pro, which has 1.6 trillion parameters, and Moonshot says it is now "the most capable model to date."

• 2.8T parameters: the first "open model in the 3T class," ahead of DeepSeek V4 Pro (1.6T)
• Price: $3 per million input tokens and $15 per million output tokens, on par with Claude Sonnet
• The previous Kimi K2.6 cost $0.95 and $4 per million tokens, so the price has risen several-fold
• Open weights are promised by July 27, 2026
• Elo 1547 on the long-horizon knowledge work test, +732 points over Kimi K2.6

How strong is it?

By Moonshot's own benchmarks, Kimi K3 beats Claude Opus 4.8 max and GPT-5.5 high but trails Claude Fable 5 and GPT-5.6 Sol. In Artificial Analysis's independent evaluation, the model took second place on the test of long work tasks, behind only Claude Fable 5.

"On our closed long-horizon knowledge work evaluation, Kimi K3 reaches an Elo of 1547, 732 points above Kimi K2.6, and is behind only Claude Fable 5,"

Artificial Analysis says in its report.

According to Artificial Analysis, Kimi K3 costs $0.94 per task, roughly the same as GPT-5.6 Sol ($1.04) and half the price of Claude Opus 4.8 ($1.80). The model also uses 21% fewer output tokens than Kimi K2.6. On Arena.ai's frontend coding leaderboard, Kimi K3 took first place, ahead of even Claude Fable 5.

Why is the price a record?

Kimi K3 is the most expensive model ever released by a Chinese lab: $3 per million input tokens and $15 per million output tokens. That matches Anthropic's Claude Sonnet line and marks a sharp jump from the previous Kimi K2.6, which cost $0.95 and $4 per million tokens.

The price rise tracks the growth in size: 2.8 trillion parameters is more than double the previous generation's trillion-parameter Kimi. For the first time, Moonshot is moving away from the low-price strategy typical of Chinese models and positioning K3 as a flagship rather than a cheap alternative to Western labs. All the figures published so far are Moonshot's self-reported benchmarks and a preliminary independent evaluation, so K3's final standing will become clearer once the weights are opened.

What this means

Chinese labs are no longer competing on price alone: Moonshot has released the largest open model and, for the first time, priced it like a Western flagship. If the weights do appear on July 27, 2026, developers will get a 2.8-trillion-parameter model they can run on their own infrastructure.

Frequently asked questions

When will the Kimi K3 open weights be released?

Moonshot AI promises to release the open weights by July 27, 2026. Until then, the model is available only through the company's website and API.

How much does Kimi K3 cost through the API?

Kimi K3 costs $3 per million input tokens and $15 per million output tokens. That matches Claude Sonnet and is the highest price among models from Chinese labs.

How does Kimi K3 differ from DeepSeek V4 Pro?

Kimi K3 is larger: 2.8 trillion parameters versus 1.6 trillion for DeepSeek V4 Pro. Size is the basis on which Moonshot calls K3 the first "open model in the 3-trillion class."