2-Trillion-Parameter Showdown: After Kimi K3 Claims the Top Spot, Moonshot AI Invites Elon Musk to "Join the Team"

A cross-platform exchange between China’s Weibo and English-language social media has pushed the large-model rivalry into full public view. What began as a leaderboard contest quickly turned into a sharper war of words after Moonshot AI’s newly released open-source model, Kimi K3, climbed to the top of a global ranking on July 20.

The timing made the moment especially charged. Just days earlier, Elon Musk had signaled that xAI was preparing a much larger model—one he suggested could outperform Kimi. That comment raised expectations and shifted the conversation from benchmark positions to a broader contest over scale, capability, and momentum.

Musk Raises the Stakes With a 2-Trillion-Parameter Claim

On the morning of July 18, Musk wrote on social media that xAI’s upcoming 2-trillion-parameter model was superior “in every way” to the company’s current 1.5-trillion-parameter version. He added that the new system was expected to complete its initial training the following week.

Musk also said the model might surpass Kimi, while maintaining reasoning speed and token efficiency close to the existing 1.5T system—identified as Grok 4.5. With Kimi K3 having just secured a leading position on the open-source leaderboard, that statement effectively escalated the comparison to an even larger scale.

Moonshot AI Fires Back With a Short, Pointed Reply

Moonshot AI did not offer a long rebuttal. Instead, it responded publicly on Weibo by tagging Musk and posting a single line: “Welcome to the 2T+ club.”

Though brief, the message carried a clear edge. Coming on the heels of Kimi K3’s rise in the rankings, the response reinforced Moonshot AI’s confident posture in the open-source model race and gave the broader debate a more attention-grabbing narrative: not just who leads today, but who belongs in the next tier of ultra-large models.

xAI’s New Model Is Still Largely Unclear

Despite the hype, many key details about xAI’s new model remain unknown. The company has not yet announced a formal release date, training specifics, or a full technical breakdown for the 2-trillion-parameter system Musk referenced.

It is also unclear whether the model in question will be introduced as Grok 4.6 or as an entirely separate product. For now, the market is left with Musk’s teaser but without the supporting disclosures needed to fully assess the claim.

Grok 4.5 Had Already Entered the Conversation

Before this latest exchange, xAI had already rolled out Grok 4.5 on July 9. According to available information, it was the company’s first model trained specifically for coding and agent-oriented tasks, with training completed jointly by xAI and Cursor.

The model has been positioned as combining frontier-level intelligence with speed and cost efficiency. Musk himself described it as an Opus-class model, underscoring xAI’s ambition to compete not just on size, but on practical performance as well.

The Next Round Has Yet to Arrive

For now, the contrast is clear: Moonshot AI’s Kimi K3 is already visible on the leaderboard, while xAI’s 2-trillion-parameter challenger is still in the pre-release stage. That leaves Kimi with the immediate spotlight, even as Musk’s comments hint at a more aggressive challenge ahead.

The larger battle—spanning parameter count, benchmark strength, inference efficiency, and industry influence—is far from settled. The next meaningful phase of this competition will likely begin only after xAI completes training and formally unveils its new model.

Until then, the “2T+ club” line may stand as the most memorable moment in a rivalry that is increasingly being fought not only in labs and rankings, but also in public, across two very different social media worlds.