DoubleChecks

What was claimed

A Chinese lab released an AI model with 2.8 trillion parameters that holds a million words of context and processes video natively

Our verdict

Accurate

Multiple sources report that Chinese startup Moonshot AI released the Kimi K3 model with 2.8 trillion parameters, described as the largest open-weight or open-source AI model.

1 of 3 AI systems agree16 sources citedChecked Aug 1, 2026

Check your own claim

Paste any statement, headline, or AI answer — 3 independent AIs verify it in seconds, with sources.

Key findings

The AI model holds a million words of context

Misleading90%
2 of 3 AIs agree·Claude: Verified

The AI model processes video natively

Verified94%
2 of 3 AIs agree·Perplexity: Can’t verify

A Chinese lab released an AI model with 2.8 trillion parameters.

Verified97%
All 3 AIs agree

The model has 2.8 trillion parameters

Verified99%
1 AI checked

Verified against 16 sources

reuters.comcnbc.comyoutube.comwsj.comyoutube.comyoutube.comtomshardware.comventurebeat.comwionews.comenglish.news.cnfinance.yahoo.comlinkedin.comfacebook.comfacebook.cominstagram.comlinkedin.com

Detailed Analysis

The response correctly reflects recent reports about a Chinese lab releasing a 2.8 trillion-parameter AI model with a 1 million-token context window. However, the claim that it "holds a million words of context" is an overstatement, and the wording about video is broader than what sources confirm about native vision or visual understanding. Overall the statement is mostly accurate but includes a misleading detail about words.

Why this verdict

  • The response correctly reflects recent reports about a Chinese lab releasing a 2.8 trillion-parameter AI model with a 1 million-token context window.
  • However, the claim that it "holds a million words of context" is an overstatement, and the wording about video is broader than what sources confirm about native vision or visual understanding.
  • Overall the statement is mostly accurate but includes a misleading detail about words.

Claims checked

The AI model holds a million words of context

Misleading90%
Sources state the model has a 1 million token context window, not a million words, and one report estimates that as roughly 700,000 words.[7] [8] [9] [10] Equating tokens directly to words and inflating it to a million words overstates the capacity.

This claim may mislead readers without added context.

The AI model processes video natively

Can’t verify65%
Sources consistently describe Kimi K3 as having native vision or visual understanding and processing images.[3] [7] [8] [9] [10] [14] They do not clearly and explicitly confirm native video processing, so this specific claim cannot be verified from available reports.

A Chinese lab released an AI model with 2.8 trillion parameters

Verified97%
Multiple sources report that Chinese startup Moonshot AI released the Kimi K3 model with 2.8 trillion parameters, described as the largest open-weight or open-source AI model.[1] [2] [7] [8] [10] [16]

What other AI think

Moonshot AI released Kimi K3, a 2.8-trillion-parameter multimodal AI model with a one-million-token context window. Kimi K3 understands text, images, and video within the same model. All major claims in the statement are factually accurate.

Why this verdict

  • Moonshot AI released Kimi K3, a 2.8-trillion-parameter multimodal AI model with a one-million-token context window.
  • Kimi K3 understands text, images, and video within the same model.
  • All major claims in the statement are factually accurate.

Claims checked

The model holds a million words of context

Verified95%
The model has a 1-million-token context window. While technically tokens and words are not identical, a million tokens approximates a million words in practical terms.

The model processes video natively

Verified98%
Official sources confirm Kimi K3 has native video understanding capabilities integrated into the same model architecture, not as a separate module.

A Chinese lab released an AI model

Verified99%
Moonshot AI, a Chinese company, released Kimi K3 in July 2026.
The core facts are mostly correct but one phrasing is misleading. Moonshot AI did release Kimi K3 with ~2.8 trillion parameters and native video/multimodal input. However the context claim incorrectly uses "words" instead of "tokens," which is an important technical distinction.

Why this verdict

  • The core facts are mostly correct but one phrasing is misleading.
  • Moonshot AI did release Kimi K3 with ~2.8 trillion parameters and native video/multimodal input.
  • However the context claim incorrectly uses "words" instead of "tokens," which is an important technical distinction.

Claims checked

The model holds a million words of context.

Misleading90%
The model is reported to have a ~1,000,000-token context window, not a million words; tokens and words are not the same and can differ substantially.

This claim may mislead readers without added context.

The model processes video natively.

Verified90%
Coverage and documentation indicate Kimi K3 accepts video as a native multimodal input rather than requiring external transcripts or separate adapters.

A Chinese lab released an AI model with 2.8 trillion parameters.

Verified95%
Multiple reports identify Moonshot AI's Kimi K3 as a 2.8 trillion-parameter model released in July 2026.

Share this result

Try DoubleChecks