GPT-4 Turbo: benchmark results
OpenAI's faster, cheaper GPT-4 tier with a 128K context. Provider: OpenAI. Released 2024-04-09. Access: API.
Aggregate Rating 1611 ± 5, rank #404 of 1656 rated models, from 1133 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AutoRace | 61 | Average Score (%, six reasoning tasks) | 100 |
| CausalT3 - L1 Association | 100 | Balanced accuracy (%; mean of Utility on valid associational | 100 |
| CausalT3 - L2 Intervention | 98 | Accuracy (%; 304 interventional vignettes, VALID/FLAWED, neu | 100 |
| CyberSecEval-3 Prompt Injection | 17 | Prompt injection success (%) | 100 |
| CyberSecEval-3 Social Engineering | 3.98 | Mean persuasion rating (1–5) | 100 |
| DS-1000 | 0.54 | Overall mean | 100 |
| EconLogicQA | 0.57 | Accuracy | 100 |
| EvalPlus (HumanEval+ & MBPP+) | 86.6 | Pass@1 avg (%) | 100 |
| EvoEval | 68.63 | Pass@1 (%) | 100 |
| EvoEval Subtle | 82 | Pass@1 (%) | 100 |
| HELM EWoK - Physical Dynamics | 91.33 | EM | 100 |
| HELM EWoK - Physical Interactions | 86.61 | EM | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-4-turbo · How It Works · Data refreshed daily, snapshot 2026-10-20.