GPT-4 Turbo Nov 2023
GPT-4 Turbo ranks first on 5 of the 26 benchmarks with five or more models scored.
30
Benchmarks with a score
5
Ranked first, out of five or more models
5
In the top three
OpenAI Nov 2023
GPT-4 Turbo
Nov 2023 snapshot Closed weights
124.8
#175 of 256
Epoch Capabilities Index
- GPQA Diamond 42.4% #180 /229
- Molecular Biology Capabilities Test 33.4% #84 /90
- GTA 46.6% #1 /15
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 | |
|---|---|---|---|---|
| AlpacaEval | 97.7% | #1 of 19 | Official leaderboard | |
| CriticEval | 7.86 | #1 of 12 | Official leaderboard | |
| AlpacaEval 2.0 | 50% | #5 of 47 | Official leaderboard | |
| FollowBench | 73.4% | #1 of 3 | Official leaderboard | |
| SEAL Multilingual | 1,034 | #9 of 19 | Official leaderboard | |
| LMArena (Elo) | 1,313 | #219 of 335 | Official leaderboard | |
| SEAL Precise Instruction Following | 79.5 | #17 of 18 | Official leaderboard |
Math and science
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| SuperCLUE-Math6 | 87.8 | #1 of 4 | Official leaderboard | |
| MATH Level 5 | 40% | #62 of 94 | Epoch AI run | |
| GPQA Diamond | 42.4% | #180 of 229 | Epoch AI run |
Coding
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| HumanEval+ | 81.7% | #10 of 46 | Official leaderboard | |
| MBPP+ | 73.3% | #8 of 35 | Official leaderboard | |
| HumanEval | 85.4% | #12 of 46 | Official leaderboard | |
| MBPP | 85.7% | #11 of 35 | Official leaderboard | |
| SWE-bench | 12.5% | #5 of 9 | Official leaderboard | |
| SEAL Coding | 992 | #16 of 26 | Official leaderboard | |
| SWE-bench Lite | 28.3% | #10 of 15 | Official leaderboard | |
| DevBench Implementation | 6.90% | #2 of 3 | Official leaderboard | |
| METR Time Horizon | 9.4 min | #22 of 31 | Official leaderboard | |
| METR Time Horizon 1.1 | 4.0 min | #20 of 26 | Official leaderboard | |
| SWE-bench Verified | 22.4% | #52 of 60 | Official leaderboard |
Agents and tools
Vision, audio and robotics
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| Image2Struct | 45.6 | #11 of 18 | Official leaderboard | |
| VHELM | 46.1 | #27 of 41 | Official leaderboard |
Safety and security
| Benchmark | Score | Rank | Source | |
|---|---|---|---|---|
| NYU CTF Bench | 7.00% | #4 of 7 | Official leaderboard | |
| Molecular Biology Capabilities Test | 33.4% | #84 of 90 | 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)