GPT-3.5 Turbo Jun 2023
GPT-3.5 Turbo places in the top three on 3 benchmarks.
26
Benchmarks with a score
0
Ranked first, out of five or more models
3
In the top three
OpenAI Jun 2023
GPT-3.5 Turbo
Jun 2023 snapshot Closed weights
106.0
#208 of 256
Epoch Capabilities Index
- Molecular Biology Capabilities Test 11.5% #90 /90
- World-Class Bio 6.60% #73 /74
- Situational Awareness Dataset (SAD) 38.4% #13 /17
Where it ranks
Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.
Knowledge and language
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| WinoGrande | 81.6% | #13 of 76 | Claim in The Falcon Series of Open Language Models, via Epoch AI | |
| BIG-Bench Hard | 61.6% | #14 of 45 | Claim in Baichuan 2: Open Large-scale Language Models, via Epoch AI | |
| AlpacaEval | 93.4% | #7 of 19 | Official leaderboard | |
| L-Eval | 71.6% | #2 of 5 | Official leaderboard | |
| CommonsenseQA 2.0 | 57% | #3 of 6 | Claim in arxiv.org, via Epoch AI | |
| AlpacaEval 2.0 | 22.7% | #26 of 47 | Official leaderboard | |
| EQ-Bench v2 | 69.3 | #53 of 85 | Official leaderboard | |
| BuzzBench | 39.0 | #14 of 20 | Official leaderboard | |
| QuALITY | 74.7% | #3 of 4 | Official leaderboard | |
| Creative Writing v3 | 42.5 | #116 of 120 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| GSM8K | 46.9% | #4 of 24 | Stanford HELM, via Epoch AI | |
| SimpleBench | 8.00% | #106 of 106 | Official leaderboard |
Coding
Professional work
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MedQA | 62.2% | #46 of 71 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Anthropic Red Team | 99.6% | #15 of 75 | Official leaderboard | |
| OR-Bench | 38.4% | #11 of 28 | Official leaderboard | |
| SuperCLUE-Safety | 83.8 | #3 of 7 | Official leaderboard | |
| AIR-Bench 2024 | 63.1% | #47 of 82 | Official leaderboard | |
| Situational Awareness Dataset (SAD) | 38.4% | #13 of 17 | Official leaderboard | |
| SimpleSafetyTests | 95.8% | #63 of 75 | Official leaderboard | |
| HELM Safety | 85.3 | #64 of 75 | Official leaderboard | |
| World-Class Bio | 6.60% | #73 of 74 | Official leaderboard | |
| Molecular Biology Capabilities Test | 11.5% | #90 of 90 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HELM Classic | 78.3 | #4 of 17 | Official leaderboard | |
| HELM Lite | 35.8 | #53 of 71 | Official leaderboard |
OpenAI's models over time
OpenAI's models in color, against every other model on the Epoch Capabilities Index, by release date.
Source: Epoch Capabilities Index, Epoch AI (CC BY 4.0)