GPT-3.5 Turbo: benchmark results
OpenAI GPT-3.5 Turbo chat model, tracked as a legacy chat/API baseline. Provider: OpenAI. Released 2022-11-30. Access: API.
Aggregate Rating 1476 ± 5, rank #865 of 1656 rated models, from 690 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CausalT3 - L2 Intervention (After Interrogation) | 87.2 | Accuracy (%; 304 interventional vignettes after a second tur | 100 |
| DoRA (Defense Documents) - Hallucination (GTE Retrieval) | 3.71 | RAGEval keypoint hallucination rate (%) of grounded answers | 100 |
| DoRA (Defense Documents) - Hallucination (Oracle Evidence) | 2.69 | RAGEval keypoint hallucination rate (%) of grounded answers | 100 |
| ES-MemEval QA (Long Context) - Conflict Detection | 21.4 | Token-level F1 (%; answers scored against human-checked refe | 100 |
| ES-MemEval QA (Long Context) - Temporal Reasoning | 22.4 | Token-level F1 (%; answers scored against human-checked refe | 100 |
| ES-MemEval QA (RAG) - User Modeling | 22 | Token-level F1 (%; answers scored against human-checked refe | 100 |
| PECC - Project Euler | 8.19 | Solve rate (%) | 100 |
| PECC - Project Euler (Story) | 6.95 | Solve rate (%) | 100 |
| RABBITS MedMCQA G2B | 97.7 | Accuracy (%) | 100 |
| RABBITS MedMCQA Original | 98.28 | Accuracy (%) | 100 |
| YapBench | 22.7 | YapIndex (lower is better) | 99 |
| RABBITS MedQA G2B | 96.03 | Accuracy (%) | 98 |
Interactive version: theaggregate.ai/model?slug=gpt-3-5-turbo · How It Works · Data refreshed daily, snapshot 2026-10-20.