Dario has increasingly turned AI existential-risk narratives into a political weapon, using fear to build public support for restrictive regulation that could undermine open-source models and protect the interests of a handful of powerful companies.
We have dumped JAX/SGLang and rewritten the entire RL and inference stack from scratch in C/C++, which took some extra time.
Training (aka pre-training) is done. Lot of cleanup was needed on the SFT data to get that mostly right. Now, 4.8 is in RL and the evals are starting to climb. We will do a soft rollout to Bot users later this week and see how it goes.
Grok 4.8 will bear no resemblance to 4.7 (thankfully). It is not yet incredible, but, this time, the foundations are built right, so I expect it to evolve to be something special.
@wickedguro I don’t really think the number of Railway outages has decreased over the past few months; if anything, the situation has gotten worse. There have been more than four incidents in the last 90 days.