NeurIPS 2025 Experts Benchmark
Models answer tasks submitted by participants at NeurIPS 2025 in San Diego. Kaggle reports an overall score and individual task outcomes.
- Made by
- Kaggle
- Released
- December 2025
- Kind
- Language reasoning
- Level
- University
- Scored
- Settled by events
- Answers
- Open-ended
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Items from
- Written by crowdworkers
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
-
Gemini 3 Pro #56 on the capability index
- Last model added
- December 2025
- Scores refreshed
- November 20th, 2026
5 models share the top score of 73.3% among models released by December 2025
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best.
The newest model with a score here came out in December 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 1
Qwen3-Coder-480B-A35B 73.3%
- 2
Gemini 3 Pro 73.3%
- 3 Claude Opus 4.5 73.3%
- 4
DeepSeek-V3.2 73.3%
- 5
Gemini 3 Flash 73.3%
- 6
Gemini 2.0 Flash Feb 2025 66.7%
- 7
DeepSeek-V3.1 66.7%
- 8
Gemini 2.5 Flash Sep 2025 66.7%
- 9 Claude Haiku 4.5 66.7%
- 10
Gemma 3 1B 53.3%
Source: Official NeurIPS 2025 Experts Benchmark leaderboard
| # | Model | Score | Setting | Released | Run on | Source |
|---|---|---|---|---|---|---|
| 1 | | 73.3% | Jul 2025 | Jan 2026 | Official leaderboard | |
| 2 | | 73.3% | Nov 2025 | Jan 2026 | Official leaderboard | |
| 3 | Claude Opus 4.5 | 73.3% | Nov 2025 | Jan 2026 | Official leaderboard | |
| 4 | | 73.3% | Dec 2025 | Jan 2026 | Official leaderboard | |
| 5 | | 73.3% | Dec 2025 | Jan 2026 | Official leaderboard | |
| 6 | | 66.7% | Feb 2025 | Jan 2026 | Official leaderboard | |
| 7 | | 66.7% | Aug 2025 | Jan 2026 | Official leaderboard | |
| 8 | | 66.7% | Sep 2025 | Mar 2026 | Official leaderboard | |
| 9 | Claude Haiku 4.5 | 66.7% | Oct 2025 | Jan 2026 | Official leaderboard | |
| 10 | | 53.3% | Mar 2025 | Jan 2026 | Official leaderboard |
Rankings by part
We read 15 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as Generate TikZ 3D Sphere in Cube, List all horse breeds and have an LLM judge and predict nvidia market value. Gemini 2.0 Flash leads 12 of them, and 3 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
Generate TikZ 3D Sphere in Cube
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8 Claude Opus 4.5 ↓5 100%
- 9
DeepSeek-V3.2 ↓5 100%
- 10
Gemini 3 Flash ↓5 100%
List all horse breeds and have an LLM judge
10 models we track have a score here.
- 1 Claude Opus 4.5 ↑2 100%
- 2
DeepSeek-V3.2 ↑2 100%
- 3
Gemini 2.0 Flash Feb 2025 ↑3 0%
- 4
Gemma 3 1B ↑6 0%
- 5
Qwen3-Coder-480B-A35B ↓4 0%
- 6
DeepSeek-V3.1 ↑1 0%
- 7
Gemini 2.5 Flash Sep 2025 ↑1 0%
- 8 Claude Haiku 4.5 ↑1 0%
- 9
Gemini 3 Pro ↓7 0%
- 10
Gemini 3 Flash ↓5 0%
predict nvidia market value
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Qwen3-Coder-480B-A35B ↓1 100%
- 3
DeepSeek-V3.1 ↑4 100%
- 4
Gemini 2.5 Flash Sep 2025 ↑4 100%
- 5
Gemini 3 Pro ↓3 100%
- 6 Claude Opus 4.5 ↓3 100%
- 7
Gemini 3 Flash ↓2 100%
- 8
Gemma 3 1B ↑2 0%
- 9 Claude Haiku 4.5 0%
- 10
DeepSeek-V3.2 ↓6 0%
Farmer's Goat Riddle
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8
DeepSeek-V3.2 ↓4 100%
- 9
Gemini 3 Flash ↓4 100%
- 10 Claude Opus 4.5 ↓7 0%
Medical Diagnosis: Empyema Necessitans
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Qwen3-Coder-480B-A35B ↓1 100%
- 3
DeepSeek-V3.1 ↑4 100%
- 4
Gemini 2.5 Flash Sep 2025 ↑4 100%
- 5 Claude Haiku 4.5 ↑4 100%
- 6
Gemini 3 Pro ↓4 100%
- 7 Claude Opus 4.5 ↓4 100%
- 8
DeepSeek-V3.2 ↓4 100%
- 9
Gemini 3 Flash ↓4 100%
- 10
Gemma 3 1B 0%
Protein Folding for KAGGLE
10 models we track have a score here.
- 1
Gemma 3 1B ↑9 100%
- 2 Claude Opus 4.5 ↑1 100%
- 3
Gemini 2.0 Flash Feb 2025 ↑3 0%
- 4
Qwen3-Coder-480B-A35B ↓3 0%
- 5
DeepSeek-V3.1 ↑2 0%
- 6
Gemini 2.5 Flash Sep 2025 ↑2 0%
- 7 Claude Haiku 4.5 ↑2 0%
- 8
Gemini 3 Pro ↓6 0%
- 9
DeepSeek-V3.2 ↓5 0%
- 10
Gemini 3 Flash ↓5 0%
Liar's Paradox
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8 Claude Opus 4.5 ↓5 100%
- 9
DeepSeek-V3.2 ↓5 100%
- 10
Gemini 3 Flash ↓5 100%
Toronto Street Grid Directions
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8 Claude Opus 4.5 ↓5 100%
- 9
DeepSeek-V3.2 ↓5 100%
- 10
Gemini 3 Flash ↓5 100%
List recent major stock splits
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 0%
- 2
Gemma 3 1B ↑8 0%
- 3
Qwen3-Coder-480B-A35B ↓2 0%
- 4
DeepSeek-V3.1 ↑3 0%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 0%
- 6 Claude Haiku 4.5 ↑3 0%
- 7
Gemini 3 Pro ↓5 0%
- 8 Claude Opus 4.5 ↓5 0%
- 9
DeepSeek-V3.2 ↓5 0%
- 10
Gemini 3 Flash ↓5 0%
pokemon birthday puns
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 0%
- 2
Gemma 3 1B ↑8 0%
- 3
Qwen3-Coder-480B-A35B ↓2 0%
- 4
DeepSeek-V3.1 ↑3 0%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 0%
- 6 Claude Haiku 4.5 ↑3 0%
- 7
Gemini 3 Pro ↓5 0%
- 8 Claude Opus 4.5 ↓5 0%
- 9
DeepSeek-V3.2 ↓5 0%
- 10
Gemini 3 Flash ↓5 0%
Name CGM and Pump Manufacturers
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8 Claude Opus 4.5 ↓5 100%
- 9
DeepSeek-V3.2 ↓5 100%
- 10
Gemini 3 Flash ↓5 100%
Countdown Math Puzzle: Make 86
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Qwen3-Coder-480B-A35B ↓1 100%
- 3 Claude Haiku 4.5 ↑6 100%
- 4
Gemini 3 Pro ↓2 100%
- 5
DeepSeek-V3.2 ↓1 100%
- 6
Gemini 3 Flash ↓1 100%
- 7
Gemma 3 1B ↑3 0%
- 8
DeepSeek-V3.1 ↓1 0%
- 9
Gemini 2.5 Flash Sep 2025 ↓1 0%
- 10 Claude Opus 4.5 ↓7 0%
Best football club in Barcelona
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8 Claude Opus 4.5 ↓5 100%
- 9
DeepSeek-V3.2 ↓5 100%
- 10
Gemini 3 Flash ↓5 100%
count English alphabets
10 models we track have a score here.
- 1
Gemini 2.0 Flash Feb 2025 ↑5 100%
- 2
Gemma 3 1B ↑8 100%
- 3
Qwen3-Coder-480B-A35B ↓2 100%
- 4
DeepSeek-V3.1 ↑3 100%
- 5
Gemini 2.5 Flash Sep 2025 ↑3 100%
- 6 Claude Haiku 4.5 ↑3 100%
- 7
Gemini 3 Pro ↓5 100%
- 8 Claude Opus 4.5 ↓5 100%
- 9
DeepSeek-V3.2 ↓5 100%
- 10
Gemini 3 Flash ↓5 100%
count 's's in mississippi
10 models we track have a score here.
- 1
Qwen3-Coder-480B-A35B 100%
- 2
DeepSeek-V3.1 ↑5 100%
- 3
Gemini 2.5 Flash Sep 2025 ↑5 100%
- 4 Claude Haiku 4.5 ↑5 100%
- 5
Gemini 3 Pro ↓3 100%
- 6 Claude Opus 4.5 ↓3 100%
- 7
DeepSeek-V3.2 ↓3 100%
- 8
Gemini 3 Flash ↓3 100%
- 9
Gemini 2.0 Flash Feb 2025 ↓3 0%
- 10
Gemma 3 1B 0%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official NeurIPS 2025 Experts Benchmark leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official NeurIPS 2025 Experts Benchmark leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Knowledge
- MMLU-Pro In use Harder 10-option version of MMLU with reasoning-focused questions across 14 subjects.
Intern-S2-397B 89.8%
- MMLU In use 57-subject multiple-choice exam questions from elementary to professional level. Scored by accuracy.
Seed1.8 92.3%
- SuperGPQA In use 26,529 graduate-level multiple-choice questions across 285 disciplines. Accuracy.
Qwen 3.6 Max (Preview) 73.9%
- WideSearch In use 200 bilingual broad info-seeking tasks: agent collects many atomic facts into a table. Scored by success rate and row/item F1.
Tencent Hy4 preview 83.9%
- BBEH micro average In use Models solve harder replacements for each of the 23 BIG-Bench Hard tasks, 4,520 examples in total. The score is the share of all examples answered correctly.
Gemma 4 31B IT 74.4%
- HLE-Diamond In use Models answer 1,000 cleaned expert questions from Humanity's Last Exam, 500 on reasoning and 500 on knowledge, closed book. Graded on exact answers. Claude Fable 5.2 77.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 11 months ago, in December 2025, for Gemini 3 Flash. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (10 entries).
- Pass Models scored 10 models we track.
- Warning Newest score The newest came 11 months ago, in December 2025, for Gemini 3 Flash.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3 Pro (#56).
- Pass Headroom The best is 73.3%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on kaggle.com: 10 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.