SciPredict leaderboard
As of 6 Nov 2026, the top score on SciPredict is 25.3% by Gemini 3 Pro (Google), ahead of Claude Opus 4.5 at 23.1%, across 15 model configurations with a published result (Scale AI SEAL, result dated 14 Jan 2026). Given an experimental setup from the literature, the model predicts the result; scored against what the paper found. SciPredict is shown for reference only and does not count toward the BenchLeader Index.
Key facts (as of 6 Nov 2026)
- Gemini 3 Pro holds the top SciPredict score at 25.3% (Scale AI SEAL, 14 Jan 2026).
- Claude Opus 4.5 is second on SciPredict at 23.1% (Scale AI SEAL, 14 Jan 2026).
- SciPredict: Predicting the outcomes of scientific experiments before they are run. Scale AI (Scale AI SEAL).
- 15 model configurations have a published SciPredict result on BenchLeader (BenchLeader, 6 Nov 2026).
- SciPredict does not feed the BenchLeader Index; it is shown for reference (BenchLeader methodology).
Predicting the outcomes of scientific experiments before they are run. Scale AI.
- Published by
- Scale AI SEAL
- Category
- Reasoning
- Index weight
- Reference only
- Models
- 15
- Data as of
- 6 Nov 2026
Scale AI SEAL Leaderboards.
What the test looks like
Given an experimental setup from the literature, the model predicts the result; scored against what the paper found.
How it is scored
Accuracy, published by Scale AI.
What to keep in mind
Tests scientific intuition; contamination is possible for published experiments.
- 1Gemini 3 Pro25.3%
- 2Claude Opus 4.523.1%
- 3Claude Sonnet 4.522.6%
- 4Gemini 3 Flash22.2%
- 5Claude Opus 4.122.2%
- 6GPT-5.220.6%
- 7o3-mini19.8%
- 8DeepSeek V319.2%
- 9Llama 3.3 70B18.2%
- 10o317.9%
- 11Gemini 2.5 Pro17.0%
- 12Qwen3 32B17.0%
- 13Qwen3 235B A22B16.6%
- 14o4-mini16.2%
- 15Llama 3.1 8B14.7%
| # | ||||
|---|---|---|---|---|
| 1 |  Gemini 3 Pro | 25.3% | 60.2 | 2026-01-14 |
| 2 |  Claude Opus 4.5 | 23.1% | 57.9 | 2026-01-14 |
| 3 | Claude Sonnet 4.5 | 22.6% | 53.7 | 2026-01-14 |
| 4 | Gemini 3 Flash | 22.2% | 56.9 | 2026-01-14 |
| 5 | Claude Opus 4.1 | 22.2% | 51.4 | 2026-01-14 |
| 6 |  GPT-5.2 | 20.6% | 57.9 | 2026-01-14 |
| 7 | o3-mini | 19.8% | 50.7 | 2026-01-14 |
| 8 |  DeepSeek V3open ↗ | 19.2% | 44.4 | 2026-01-14 |
| 9 |  Llama 3.3 70Bopen ↗ | 18.2% | 40.0 | 2026-01-14 |
| 10 | o3 | 17.9% | 61.0 | 2026-01-14 |
| 11 | Gemini 2.5 Pro | 17.0% | 54.5 | 2026-01-14 |
| 12 |  Qwen3 32Bopen ↗ | 17.0% | 51.4 | 2026-01-14 |
| 13 | Qwen3 235B A22Bopen ↗ | 16.6% | 50.0 | 2026-01-14 |
| 14 | o4-mini | 16.2% | 56.9 | 2026-01-14 |
| 15 | Llama 3.1 8Bopen ↗ | 14.7% | 33.7 | 2026-01-14 |
More reasoning benchmarks · All LLM benchmarks
Related benchmarks
Cite as: BenchLeader, “SciPredict leaderboard”, data as of 6 Nov 2026, https://www.benchleader.com/benchmarks/scale_scipredict.
SciPredict: questions
- What does SciPredict measure?
- Given an experimental setup from the literature, the model predicts the result; scored against what the paper found. Scores are reported in percent of tasks solved; higher is better.
- What is the top score on SciPredict?
- As of 6 Nov 2026, the highest score on SciPredict is 25.3% by Gemini 3 Pro (result dated 14 Jan 2026), ahead of Claude Opus 4.5 at 23.1%. Scores come from Scale AI SEAL and are re-read daily.
- Which AI model leads SciPredict?
- Gemini 3 Pro leads SciPredict with 25.3% as of 6 Nov 2026 (result dated 14 Jan 2026), ahead of Claude Opus 4.5 at 23.1%.
- How many models have SciPredict results?
- 15 model configurations have a SciPredict result on BenchLeader, all taken from Scale AI SEAL.
- Who runs SciPredict and how often is it updated?
- SciPredict is published by Scale AI SEAL. BenchLeader re-reads the published results once a day and records the date each result was published.
- Does SciPredict count toward the BenchLeader Index?
- No. SciPredict is shown for reference but left out of the composite index.