Grok 4.7 delayed a few more days for training refinements
Elon Musk says the AI model needs adjustment to its reinforcement learning penalty system.
Grok 4.7, xAI's large language model, requires additional development time before release. "We might have penalized response length too much (or something) in RL, as it still gives up on hard tasks (that it can do!) too early and isn't yet sufficiently rigorous in checking its work," Musk wrote on X.
The delay signals that the team is refining how the model balances response brevity with thoroughness. The issue appears tied to the reinforcement learning training process, where the model may be cutting corners on complex problems it is technically capable of solving, rather than persisting to verify solutions.
No new release date was specified, only that the model needs "a few more days to cook."
"@farzyness Grok 4.7 needs a few more days to cook. We might have penalized response length too much (or something) in RL, as it still gives up on hard tasks (that it can do!) too early and isn’t yet sufficiently rigorous in checking its work."— X · @elonmusk