GPT-3.5 Turbo Instruct: benchmark results
Provider: OpenAI. Released 2022-11-30. Access: API.
Aggregate Rating 1520 ± 39, rank #1993 of 10552 rated models, from 16 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| AI Chess Leaderboard (Continuation) | 1447 | Elo | 92.4 |
| Chess Bench LLM | 1643 | Lichess-Classical-calibrated Glicko-2 rating (native signed | 85.4 |
| NCKI kspine_v3 | 68.24 | NCKI kspine_v3 index | 38.9 |
| AI Chess Leaderboard (Reasoning) | 573 | Elo | 26.3 |
| NCRI 15.2 | 49.65 | NCRI 15.2 index | 24.6 |
| AgentDrive - Physics | 27.5 | Accuracy (%) | 22 |
| AgentDrive - Scenario | 82.5 | Accuracy (%) | 12 |
| METR Benchmark | 0.01 | 50% Time Horizon (hours) | 8 |
| AgentDrive - Comparative | 52.5 | Accuracy (%) | 7 |
| AgentDrive - Hybrid | 5 | Accuracy (%) | 7 |
| AgentDrive | 34.5 | Accuracy (%) | 6 |
| TasteVal - Performance Multiplier | 0.11 | Performance multiplier against the expert baseline: test sco | 4.3 |
Interactive version: theaggregate.ai/model?slug=gpt-3-5-turbo-instruct · How It Works · Data refreshed daily, snapshot 2026-11-16.