Benchmark
MLE-Bench Lite
Agentic benchmarks measure how well models carry out tasks that take several steps. Scores are percentages. Higher is better.
- Models measured
- 18
- Scores recorded
- 33
- Sources
- 3
- IQ weight
- 2.65
Score distribution
- 59%
- Mean
- 66.6%
- Median
- 22%
- Std dev
- 18.6%
- 10th percentile
- 78.8%
- 90th percentile
- 4.6%
- Min
This benchmark has a weight of 2.65 for chat models, measured over 18 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 | Gemini 3 Pro PreviewFamou-Agent 2.0· MLE-bench official data | 80.3% |
| 2 | Gemini 3 Pro PreviewMLEvolve· MLE-bench official data | 80.3% |
| 3 | Gemini 3 Pro PreviewPiEvolve (Fractal AI Research)· MLE-bench official data | 80.3% |
| 4 | Gemini 3 Pro PreviewCAIR MARS+· MLE-bench official data | 78.8% |
| 5 | Claude Opus 4.6AIBuildAI· MLE-bench official data | 77.3% |
| 6 | Gemini 3 FlashLoongFlow· MLE-bench official data | 77.3% |
| 7 | DeepSeek V3.2 SpecialeML-Master 2.0· MLE-bench official data | 75.8% |
| 8 | Gemini 2.5 ProFamou-Agent 2.0· MLE-bench official data | 75.8% |
| 9 | Claude Opus 4.6MiniMax M2.7 official benchmark overview · exact harness/version not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.7% |
| 10 | Gemini 3 Pro PreviewCAIR MARS· MLE-bench official data | 74.2% |
| 11 | Claude Sonnet 4.6MiniMax M2.7 official benchmark overview · exact harness/version not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 72.7% |
| 12 | GPT 5.4MiniMax M2.7 official benchmark overview · exact harness/version not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 71.2% |
| 13 | Gemini 2.5 ProCAIR MLE-STAR-Pro-1.5· MLE-bench official data | 68.2% |
| 14 | Gemini 3 Pro PreviewLeeroo· MLE-bench official data | 68.2% |
| 15 | GPT 5R&D-Agent· MLE-bench official data | 68.2% |
| 16 | Gemini 2.5 ProCAIR MLE-STAR-Pro-1.0· MLE-bench official data | 66.7% |
| 17 | MiniMax M2.7· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 66.6% |
| 18 | Gemini 3.1 Pro PreviewMiniMax M2.7 official benchmark overview · exact harness/version not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 66.6% |
| 19 | MiniMax M2.7MiniMax M2.7 official benchmark overview · exact harness/version not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 66.6% |
| 20 | GPT-5 CodexThesis· MLE-bench official data | 65.2% |
| 21 | GPT 5Operand ensemble· MLE-bench official data | 63.6% |
| 22 | DeepSeek R1InternAgent· MLE-bench official data | 62.1% |
| 23 | Gemini 2.5 ProFamou-Agent· MLE-bench official data | 62.1% |
| 24 | OpenAI o3AIRA-dojo· MLE-bench official data | 55% |
| 25 | MiniMax M2.5MiniMax M2.7 official benchmark overview · exact harness/version not stated· Reported by MiniMax· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 51.5% |
| 26 | DeepSeek R1ML-Master· MLE-bench official data | 48.5% |
| 27 | O1 PreviewR&D-Agent· MLE-bench official data | 48.2% |
| 28 | O1 PreviewAIDE· MLE-bench official data | 35.9% |
| 29 | Claude 3.5 SonnetAIDE· MLE-bench official data | 19.7% |
| 30 | GPT 4oAIDE· MLE-bench official data | 18.6% |
| 31 | GPT 4oOpenHands· MLE-bench official data | 12.1% |
| 32 | Llama 3.1 405B InstructAIDE· MLE-bench official data | 10.2% |
| 33 | GPT 4oMLAB· MLE-bench official data | 4.6% |
Data credits: 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. · MLE-bench (OpenAI), MIT (licence) · Data by LLM Stats