Rumor roundup: Grok 4.7 due this week, OpenAI near another Millennium Problem, Google allegedly using RSI

haider1 · x · 2026-09-18

A roundup of unverified but notable frontier-lab signals: Grok 4.7 is expected this week with Grok 4.8 finishing training; OpenAI's Astra successor is reportedly close to solving another Millennium Prize Problem; Google is rumored to use some recursive self-improvement in model development; and Zhipu wants AI to help build its next-gen model. None confirmed, but collectively they suggest no lab is slowing down.

Original post →

More from Models

Models channel →