Token Gremlin
@TokenGremlin · 14 авг. 2026 г.
Quick clarification on the new Grok 4.6 ARC-AGI-3 result, because the chart can be misleading if you don’t know the harness story.
ARC Prize’s standard harness made GPT-5.6 Sol look much weaker than it actually is because it was effectively doing this:
Sol: “I learned that the
ARC Prize@arcprize· 14 авг. 2026 г.Grok 4.6 from @SpaceXAI on ARC-AGI (Verified):
- ARC-AGI-1: 87.5%, $0.30/task
- ARC-AGI-2: 67.1%, $0.76/task
- ARC-AGI-3: 2.11%, $5.6K
On ARC-AGI-3, Grok 4.6 with xhigh reasoning scored comparably to GPT-5.6 Sol with high reasoning, but cost $5.6K versus Sol's $15.2K.
Elon Musk
@elonmusk
Grok performs far better with its Build hardness
05:48 · 14 августа 2026 г. · 116,1 тыс. просмотров
156
114
1,6 тыс.