© 2026 Improve the News Foundation.
All rights reserved.
Version 7.7.2
Line chart tracking how leading AI model families moved across crowd-ranked thinking benchmarks from 2024 to 2026.
| Claude 3.5 -> Opus 4.6 | May 2024: #4 | Aug 2024: #2 | Nov 2024: #1 | Feb 2025: #2 | May 2025: #2 | Aug 2025: #2 | Dec 2025: #2 | Apr 2026: #1 |
|---|---|---|---|---|---|---|---|---|
| Gemini 1.5 -> 3 Pro | May 2024: #5 | Aug 2024: #4 | Nov 2024: #2 | Feb 2025: #1 | May 2025: #1 | Aug 2025: #1 | Dec 2025: #1 | Apr 2026: #2 |
| GPT 4o -> GPT-5 series | May 2024: #1 | Aug 2024: #1 | Nov 2024: #3 | Feb 2025: #4 | May 2025: #3 | Aug 2025: #4 | Dec 2025: #4 | Apr 2026: #5 |
| DeepSeek R1 / V3 | May 2024: Not ranked | Aug 2024: Not ranked | Nov 2024: Not ranked | Feb 2025: #3 | May 2025: #4 | Aug 2025: #3 | Dec 2025: #5 | Apr 2026: #6 |
| Grok 2 -> 4 | May 2024: Not ranked | Aug 2024: Not ranked | Nov 2024: #7 | Feb 2025: #6 | May 2025: #5 | Aug 2025: #5 | Dec 2025: #3 | Apr 2026: #4 |
| Llama 3 -> 4 | May 2024: #5 | Aug 2024: #5 | Nov 2024: #6 | Feb 2025: #5 | May 2025: #6 | Aug 2025: #6 | Dec 2025: #6 | Apr 2026: #7 |
Source: Arena Leaderboard. Models grouped by family for mock chart standardization.