GPT-3.5 Turbo Nov 2023
GPT-3.5 Turbo ranks first on ANLI.
28
Benchmarks with a score
1
Ranked first, out of five or more models
2
In the top three
OpenAI Nov 2023
GPT-3.5 Turbo
Nov 2023 snapshot Closed weights
112.9
#200 of 256
Epoch Capabilities Index
- GPQA Diamond 28% #212 /229
- SimpleQA Verified 11.2% #45 /56
- NYU CTF Bench 4.00% #6 /7
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 | |
|---|---|---|---|---|
| ANLI | 58.1% | #1 of 11 | Claim in Phi-3 Technical Report: A Highly Capable Language Model L…, via Epoch AI | |
| ARC-Challenge | 87.4% | #7 of 75 | Claim in Phi-3 Technical Report: A Highly Capable Language Model L…, via Epoch AI | |
| OpenBookQA | 86% | #4 of 41 | Claim in Phi-3 Technical Report: A Highly Capable Language Model L…, via Epoch AI | |
| TriviaQA | 85.8% | #6 of 38 | Claim in Phi-3 Technical Report: A Highly Capable Language Model L…, via Epoch AI | |
| MMLU | 71.4% | #22 of 85 | Claim in Phi-3 Technical Report: A Highly Capable Language Model L…, via Epoch AI | |
| EQ-Bench v2 | 71.7 | #49 of 85 | Official leaderboard | |
| AlpacaEval 2.0 | 19.3% | #29 of 47 | Official leaderboard | |
| FollowBench | 66.2% | #2 of 3 | Official leaderboard | |
| WinoGrande | 68.8% | #51 of 76 | Claim in Phi-3 Technical Report: A Highly Capable Language Model L…, via Epoch AI | |
| AlpacaEval | 86.3% | #13 of 19 | Official leaderboard | |
| SimpleQA Verified | 11.2% | #45 of 56 | Official leaderboard | |
| LMArena (Elo) | 1,204 | #282 of 335 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| MATH Level 5 | 15.9% | #80 of 94 | Epoch AI run | |
| GPQA Diamond | 28% | #212 of 229 | Epoch AI run | |
| DTBench | 48.5% | #171 of 180 | Official leaderboard |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HumanEval | 76.8% | #17 of 46 | Official leaderboard | |
| MBPP+ | 69.7% | #13 of 35 | Official leaderboard | |
| HumanEval+ | 70.7% | #18 of 46 | Official leaderboard | |
| MBPP | 82.5% | #15 of 35 | Official leaderboard | |
| CodeEditorBench | 72.4% | #6 of 7 | Official leaderboard |
Agents and tools
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| T-Eval | 84% | #2 of 12 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Anthropic Red Team | 98% | #57 of 75 | Official leaderboard | |
| AIR-Bench 2024 | 52.5% | #63 of 82 | Official leaderboard | |
| NYU CTF Bench | 4.00% | #6 of 7 | Official leaderboard | |
| SimpleSafetyTests | 92.3% | #68 of 75 | Official leaderboard | |
| HELM Safety | 83.5 | #69 of 75 | Official leaderboard |
Other
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Conceptual Reasoning Index | 12.7 | #143 of 151 | Official leaderboard | |
| ACCoRD | 20.2 | #147 of 152 | 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)