xAI’s Next Model Outperforms Grok 4.5 in Every Metric — and It Hasn’t Finished Training Yet
Elon Musk confirmed xAI’s next model outperforms Grok 4.5 in every metric, even though initial training is not yet complete.
xAI’s Next Model Outperforms Grok 4.5 Before Training Ends
Elon Musk recently confirmed a major milestone for his AI lab. Specifically, xAI’s next model outperforms Grok 4.5 in every metric on internal benchmarks. Furthermore, this massive 2-trillion-parameter system has not even finished initial training.
Musk expects the training phase to wrap up within the week. This is not a speculative leak or a vague roadmap promise. Instead, it is a public confirmation from xAI’s founder that the next generation of Grok is already more capable.
Key Details About the New 2-Trillion-Parameter System
The incoming model features 2 trillion parameters. Consequently, it enters a scale category few systems have publicly reached. Moreover, Musk noted it surpasses Grok 4.5 on every measured dimension so far.
He also suggested it may challenge Kimi K3. That open model from Moonshot AI has dominated leaderboards since its launch. For more context on this competitor, read our coverage on how the Kimi K3 Launch Imminent: Moonshot AI’s 2.5T-Parameter Model Set to Drop on July 15.
However, the most critical point for developers involves efficiency. Speed and token usage remain close to parity with the previous version. Therefore, the next Grok aims to be smarter without becoming slower or more expensive.
Why AI Model Efficiency Matters for This Upgrade
Grok 4.5 previously broke industry patterns by delivering frontier intelligence at low cost. Usually, more capability means higher latency and greater token usage. Scaling further without degrading efficiency is genuinely difficult.
It requires architectural discipline rather than just more compute. Fortunately, xAI signals that speed remains consistent. This suggests a targeted intelligence upgrade rather than brute-force expansion.
Benchmarking Against Kimi K3
Musk’s comparison to Kimi K3 sets a specific performance target. Since its release, Kimi K3 has been one of the most impressive open-weight models ever released. It features native vision and deep reasoning capabilities.
Additionally, it possesses enough coding capability to top LMArena’s Code Arena leaderboard. Positioning xAI’s new model against this competitor frames the next phase of AI competition clearly. The race now focuses on hitting high capability ceilings while keeping models practical at scale.
What Finishing Training Means for Release Timelines
Finishing initial training does not mean a public release is imminent. Post-training work like alignment and safety evaluation typically follows. This process can take weeks or months to complete properly.
However, completing initial training is a genuine milestone. Internal benchmarks already show superiority before this process ends. Thus, the model likely has headroom to improve further through post-training optimization.
For context on xAI’s roadmap, the SuperGrok Heavy subscription tier recently bundled X Premium+. This signals active repositioning ahead of a significant model upgrade.
Smarter AI Architecture, Not Just Bigger
Musk emphasized that the model is getting smarter while staying efficient. This distinction matters because parameter count does not equal capability. The current competitive frontier combines peak intelligence with practical deployability.
Whether independent benchmarking validates these claims remains to be seen. Nevertheless, the specific comparisons make this a concrete announcement. Finally, the coming weeks will clarify how much of this holds true.
Frequently Asked Questions
Has xAI’s next model finished training?
No, Elon Musk stated that initial training is not yet complete. However, he expects it to finish within the week, and internal benchmarks already show it surpassing Grok 4.5.
How does the new Grok compare to Kimi K3?
Musk specifically positioned the new model as a challenger to Kimi K3. This suggests xAI is targeting similar benchmarks in coding, vision, and reasoning where Kimi K3 currently leads.
Will the new Grok model be slower or more expensive?
Current indications suggest otherwise. Musk noted that speed and token usage remain close to parity with Grok 4.5, prioritizing AI model efficiency alongside raw intelligence gains.