🌐 LLM Leaderboard Update 🌐 New ARC-AGI benchmarks are here! 🔥 #Gemini3DeepThink dominates #ARCAGI1 at 96.0%, while #GPT52Refine takes #ARCAGI2 at 84.6%. #ClaudeOpus46 close behind in both! === ARC-AGI-1 Leaderboard === 1. Gemini 3 Deep Think (2/26) - 96.0% 2. GPT-5.2 (Refine.) - 94.5% 3. Claude Opus 4.6 (120K, High) - 94.0% 4. Claude Opus 4.6 (120K, Max) - 93.0% 5. Claude Opus 4.6 (120K, Medium) - 92.0% 6. GPT-5.2 Pro (X-High) - 90.5% 7. Gemini 3 Deep Think (Preview) ² - 87.5% 8. GPT-5.2 (X-High) - 86.2% 9. Claude Opus 4.6 (120K, Low) - 86.0% 10. GPT-5.2 Pro (High) - 85.7% 11. Gemini 3 Flash Preview (High) - 84.7% 12. GPT-5.2 Pro (Medium) - 81.2% 13. Opus 4.5 (Thinking, 64K) - 80.0% 14. Grok 4 (Refine.) - 79.6% 15. GPT-5.2 (High) - 78.7% 16. Opus 4.5 (Thinking, 32K) - 75.8% 17. Gemini 3 Pro - 75.0% 18. GPT-5.1 (Thinking, High) - 72.8% 19. GPT-5.2 (Medium) - 72.7% 20. Opus 4.5 (Thinking, 16K) - 72.0% === ARC-AGI-2 Leaderboard === 1. Gemini 3 Deep Think (2/26) - 84.6% 2. GPT-5.2 (Refine.) - 72.9% 3. Claude Opus 4.6 (120K, High) - 69.2% 4. Claude Opus 4.6 (120K, Max) - 68.8% 5. Claude Opus 4.6 (120K, Medium) - 66.3% 6. Claude Opus 4.6 (120K, Low) - 64.6% 7. GPT-5.2 Pro (High) - 54.2% 8. Gemini 3 Pro (Refine.) - 54.0% 9. GPT-5.2 (X-High) - 52.9% 10. Gemini 3 Deep Think (Preview) ² - 45.1% 11. GPT-5.2 (High) - 43.3% 12. GPT-5.2 Pro (Medium) - 38.5% 13. Opus 4.5 (Thinking, 64K) - 37.6% 14. Gemini 3 Flash Preview (High) - 33.6% 15. Gemini 3 Pro - 31.1% 16. Grok 4 (Refine.) - 29.4% 17. NVARC - 27.6% 18. GPT-5.2 (Medium) - 26.7% 19. Opus 4.5 (Thinking, 16K) - 22.8% 20. GPT-5 Pro - 18.3% #ARCAGI1 #ARCAGI2 #ai #LLM