Benchmark
GDPval
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 11
- Scores recorded
- 19
- Sources
- 2
- IQ weight
- 3.09
Score distribution
- 69.1%
- Mean
- 70.9%
- Median
- 13.5%
- Std dev
- 46.7%
- 10th percentile
- 83%
- 90th percentile
- 45.8%
- Min
This benchmark has a weight of 3.09 for chat models, measured over 10 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 87.9% |
| 2 | GPT 5.5wins or ties · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 84.9% |
| 3 | GPT 5.4wins or ties· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 83% |
| 4 | Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 82.7% |
| 5 | GPT-5.5 Prowins or ties · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82.3% |
| 6 | GPT-5.4 Prowins or ties· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 82% |
| 7 | GPT-5.2 Prowins or ties· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 74.1% |
| 8 | GPT-5.2 Prowins or ties · ties allowed · ChatGPT heavy· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 74.1% |
| 9 | GPT 5.2wins or ties· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70.9% |
| 10 | GPT 5.2wins or ties · ties allowed · ChatGPT heavy· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70.9% |
| 11 | GPT-5.3 Codexwins or ties· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70.9% |
| 12 | GPT-5.3 Codexwins or ties · custom skills · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 70.9% |
| 13 | GPT-5.2 Prowins · no ties · ChatGPT heavy· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 67.6% |
| 14 | GPT 5.2wins · no ties · ChatGPT heavy· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 61% |
| 15 | GPT-5.2 Proclear wins · ties allowed · ChatGPT heavy· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 60% |
| 16 | GPT 5.2clear wins · ties allowed · ChatGPT heavy· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 49.8% |
| 17 | Nemotron 3 Ultra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 46.7% |
| 18 | Nemotron 3 UltraNeMo Evaluator SDK · official table· Reported by NVIDIA· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 46.7% |
| 19 | Step 3.7 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 45.8% |
Data credits: Data by LLM Stats · BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links.