Rankings / OpenAI
GPT 4o (2024-11-20)
released 2024-11-20
BenchAtlas Index
as of 2026-07-14
high effort
rank #398
8 families · 4 categories · medium
base configuration
rank #409
6 families · 3 categories · medium
Benchmark evidence
20 results
external indices
| ECI | 129.3 points | independent | Epoch AI Benchmarking Hub |
knowledge science
| GPQA Diamond implementation=vals-ai | high effort | 53.8 % | independent | Vals AI |
| GPQA Diamond | 47.9 % | independent | Epoch AI Benchmarking Hub 2025-02-05 | |
| MMLU-Pro implementation=vals-ai | high effort | 72.6 % | independent | Vals AI |
multimodal
| MMMU implementation=vals-ai | high effort | 62.2 % | independent | Vals AI |
professional
| CaseLaw | high effort | 59.7 % | independent | Vals AI |
| CorpFin | high effort | 45.9 % | independent | Vals AI |
| LegalBench | high effort | 82.2 % | independent | Vals AI |
| TaxEval | high effort | 74.5 % | independent | Vals AI |
reasoning math
| AIME implementation=vals-ai | high effort | 11.9 % | independent | Vals AI |
| ARC-AGI-1older version split=semi_private · model_type=Base LLM | 4.5 % | independent | ARC Prize Leaderboard | |
| ARC-AGI-2 split=public_eval · model_type=Base LLM | 0.1 usd_per_task | independent | ARC Prize Leaderboard | |
| ARC-AGI-2 split=semi_private · model_type=Base LLM | 0.1 usd_per_task | independent | ARC Prize Leaderboard | |
| ARC-AGI-1older version split=semi_private · model_type=Base LLM | 0.1 usd_per_task | independent | ARC Prize Leaderboard | |
| ARC-AGI-2 split=semi_private · model_type=Base LLM | 0.0 % | independent | ARC Prize Leaderboard | |
| ARC-AGI-2 split=public_eval · model_type=Base LLM | 0.0 % | independent | ARC Prize Leaderboard | |
| FrontierMath | 0.3 % | independent | Epoch AI Benchmarking Hub 2025-03-06 | |
| MATH-500 implementation=vals-ai | high effort | 74.0 % | independent | Vals AI |
| MATH Level 5 | 49.8 % | independent | Epoch AI Benchmarking Hub 2025-02-05 | |
| OTIS Mock AIME 2024–2025 | 6.3 % | independent | Epoch AI Benchmarking Hub 2025-02-25 |
Agent + model results
systems, not bare-model scores
| agent + model Epoch Inspect harness + GPT 4o (2024-11-20) | SWE-bench Verified | 31.0 % | independent | Epoch AI Benchmarking Hub |
These scores measure the whole agent system (scaffold, tools, budgets) — they are never merged into the bare model’s numbers.