🚨 xAI: GROK’S REASONING POWER HAS GOTTEN A MASSIVE COMPUTE BOOST “It’s important to realize there are two types of training compute. One is the pre-training compute—that’s from Grok 2 to Grok 3. But from Grok 3 to Grok 4, we’re actually putting a lot of compute into reasoning.” Source: @xai @elonmusk
Mario Nawfal
Mario Nawfal10.7. klo 12.19
🚨GROK SCORES 26.9% ON "HUMANITY'S LAST EXAM" WITHOUT ANY TOOLS The scaling graph tells the story: more compute = better performance. Grok crushed over a quarter of the world's hardest academic benchmark using pure reasoning alone. No calculators, no external help. Just raw AI brainpower tackling 2,500 questions across every field of human knowledge. Most humans would fail this test even WITH tools. Grok's doing it blindfolded. Source: @xai @elonmusk
54,67K