Elon Musk Announces: 2-Trillion Parameter Grok 4.6 to Finish Training Next Week, May Surpass Kimi K3

TradingKey - On July 18, Eastern Time, SpaceX ( SPCX) and Tesla ( TSLA) CEO Elon Musk revealed on social platform X that training for the next-generation large model Grok 4.6, developed by his AI company xAI, has entered its final stages and initial training is expected to be completed next week. (Note: Musk's original words on July 18 were "next week"; based on the post date of July 20, this refers to this week.)

[Source: X]

According to Musk, this new model with 2 trillion parameters is "superior in all aspects" to the current 1.5-trillion-parameter Grok 4.5 version, while its inference speed and token efficiency will remain largely on par with the previous generation. In a reply to a user, he stated that this model "has the potential to surpass" Kimi K3, recently released by Moonshot AI.

These remarks come shortly after Chinese AI startup Moonshot AI released Kimi K3. On July 17, Moonshot AI officially launched its next-generation large model, Kimi K3. With a parameter scale of 2.8 trillion, it is currently the world's largest open-source model by parameter count. On the day of the release, Musk commented "impressive" in the comments section of a related review report.

Following its release, Kimi K3 delivered an outstanding performance on third-party evaluation platforms. On the Arena AI front-end coding leaderboard, K3 topped the global rankings with a score of 1679, surpassing Anthropic's Claude Fable 5 and OpenAI's GPT-5.6 Sol. Meanwhile, Moonshot AI officially admitted that K3's overall user experience still lags significantly behind these two top-tier closed-source models.

Kimi's official account responded to Musk's remarks, posting on the social platform Weibo: "@ElonMusk Welcome to the '2 Trillion+' club." It is reported that the full model weights of Kimi K3 will be open to developers worldwide before July 27.

[Source: Weibo]

Currently, xAI has not yet announced the official launch date, full training scheme, or technical evaluation metrics for Grok 4.6. Looking back at the development process, xAI completed the training of its 1.5-trillion-parameter base model V9-Medium in May 2026, which is regarded as the core foundation of Grok 4.5.

In terms of computing power support, the Grok 4 series models rely on the ultra-large-scale AI computing cluster code-named "Colossus" for training. In January 2026, xAI announced that Colossus 2 had become operational, equipped with approximately 555,000 GB200 and GB300 series GPUs at launch, equivalent to the computing power of roughly 1.4 million H100 GPUs. xAI previously revealed that the ultimate goal for Colossus 2 is a capacity of 2 gigawatts. In June 2026, it plans to add more than 400 megawatts of computing capacity, adding over 220,000 GB300 chips, with a final target of deploying 1 million GPUs.

Industry analysis points out that while the total number of parameters in a large language model is an important reference for measuring model scale, its actual comprehensive capabilities also depend on multiple factors, such as training data quality, algorithmic frameworks, inference architectures, computing power supply, and post-training optimization. Whether xAI can maintain the operational efficiency of previous generations while expanding its model scale will be key to whether Grok 4.6 can pose a substantial challenge to Kimi K3.