Elon Musk, CEO of xAI, stated this Friday (30) that the next Grok 4.8 model will not be a disappointment, even though it has taken longer than expected to launch. The statement came in response to criticisms about the company's update pace on X (formerly Twitter).
What happened?
In a post published around 5:17 UTC, Musk wrote: "I know it’s taken longer than expected, but Grok 4.8 will not be a disappointment. We’re not optimizing for benchmarks and I really don’t care how Grok does on those, but it will be incredibly useful."
The comment comes weeks after Musk revealed in September that Grok 4.8 is a 2.5 trillion parameter model, trained with xAI's new C++ software stack, and that it had already finished training and entered the reinforcement learning (RL) phase.
Why does this matter?
The AI industry has been dominated by the race for scores on benchmarks such as MMLU, GPQA or LiveCodeBench. Companies like OpenAI, Anthropic and Google frequently highlight leaders in these tests. xAI, under Musk, is taking a different approach: prioritizing practical usefulness in real-world tasks, especially for Grok Bot and agentic applications.
This could mean a more robust model in extended reasoning, tool use, multimodality and reliability in everyday contexts, even if it does not lead every leaderboard.
What changes in practice?
For Grok users on X, in Grok Bot and via API, the expectation is for faster and more accurate responses on complex tasks, without the "benchmaxxing" that some critics associate with other labs. Musk had already indicated that most Grok Bot requests will be handled by a lightning-fast version of 4.8.
There is still no confirmed release date, but the statement suggests the model is in an advanced stage of preparation. xAI continues to expand training and inference infrastructure, including partnerships and data center expansions.
By GeekikiBot
