GPT-4 (0613): benchmark results
June 13, 2023 GPT-4 API snapshot, kept separate from other GPT-4 rows when sources report it explicitly. Provider: OpenAI. Released 2023-06-13. Access: API.
Aggregate Rating 1586 ± 8, rank #491 of 1656 rated models, from 616 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| CanItEdit (Descriptive) | 63.33 | pass@1 (%; share of the 105 hand-written Python editing prob | 100 |
| CanItEdit (Lazy) | 51.95 | pass@1 (%; share of the 105 hand-written Python editing prob | 100 |
| CodeEditorBench Plus - Code Debug (Three-shot) | 34.5 | pass@1 (%, greedy decoding, three-shot, CodeEditorBench_Plus | 100 |
| CodeEditorBench Plus - Code Debug (Zero-shot) | 31.6 | pass@1 (%, greedy decoding, zero-shot, CodeEditorBench_Plus) | 100 |
| CodeEditorBench Plus - Code Requirement Switch (Three-shot) | 30.3 | pass@1 (%, greedy decoding, three-shot, CodeEditorBench_Plus | 100 |
| CodeEditorBench Plus - Code Requirement Switch (Zero-shot) | 26.4 | pass@1 (%, greedy decoding, zero-shot, CodeEditorBench_Plus) | 100 |
| CodeEditorBench Plus - Code Translate (Three-shot) | 51.7 | pass@1 (%, greedy decoding, three-shot, CodeEditorBench_Plus | 100 |
| CommonGen | 45.11 | Overall (%) | 100 |
| FastEval | 77.78 | Total Score | 100 |
| HELM CLEVA - Chinese Cleva | 96.67 | Mean win rate | 100 |
| HELM CLEVA - Cleva Classical Chinese Understanding | 78.25 | EM | 100 |
| HELM CLEVA - Cleva Commonsense Reasoning | 69.44 | EM | 100 |
Interactive version: theaggregate.ai/model?slug=gpt-4-0613 · How It Works · Data refreshed daily, snapshot 2026-10-20.