NPHardEval
Models solve algorithmic decision and optimization problems that vary in computational complexity.
The leaderboard reports weighted accuracy by problem type.
- Made by
- NPHardEval team
- Released
- December 2023
- Kind
- Language reasoning
- Level
- University
- Scored
- Checked automatically
- Answers
- Short answer
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- New items on a schedule
- Run by
- Anyone
- In views
- Kept fresh
- Strongest model tested
- GPT-4 Turbo Apr 2024 #172 on the capability index
- Last model added
- April 2024
- Scores refreshed
- November 20th, 2026
GPT-4 Turbo leads with 37.7% among models released by April 2024 GPT-4 Turbo leads with 37.7% among models released by April 2024
Best score per entry from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Names in gray aren't in our model list: fine-tunes, agent systems, products and older models. They don't count toward leaders. This view shows only models we track, each at its best.
The newest model with a score here came out in April 2024, 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 GPT-4 Turbo Apr 2024 37.7%
- 2 Claude 2 26.4%
- 3 GPT-3.5 Turbo Jan 2024 25.9%
- 4 Claude Instant 23.9%
- 5
Qwen-14B 22.3%
- 6 01-ai/Yi-34B-Chat 18.9%
- 7
Mistral-7B-Instruct-v0.1 18.2%
- 8 PaLM-2 16.5%
- 9
Phi-2 9.21%
- 10
Vicuna-13B-v1.3 7.56%
- 11
Phi-1.5 0.22%
- 12 mosaicml/mpt-30b-instruct 0.02%
Source: Official NPHardEval leaderboard
- 1 GPT-4 Turbo Apr 2024 37.7%
- 2 Claude 2 26.4%
- 3 GPT-3.5 Turbo Jan 2024 25.9%
- 4 Claude Instant 23.9%
- 5
Qwen-14B 22.3%
- 6
Mistral-7B-Instruct-v0.1 18.2%
- 7
Phi-2 9.21%
- 8
Vicuna-13B-v1.3 7.56%
- 9
Phi-1.5 0.22%
Source: Official NPHardEval leaderboard
| # | Entry | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-4 Turbo Apr 2024 | 37.7% | Apr 2024 | Official leaderboard | ||
| 2 | Claude 2 | 26.4% | Jul 2023 | Official leaderboard | ||
| 3 | GPT-3.5 Turbo Jan 2024 | 25.9% | Jan 2024 | Official leaderboard | ||
| 4 | Claude Instant | 23.9% | Aug 2023 | Official leaderboard | ||
| 5 | | 22.3% | Sep 2023 | huggingface.co | Official leaderboard | |
| 6 | 01-ai/Yi-34B-Chat open weights, 34.4B | 18.9% | huggingface.co | Official leaderboard | ||
| 7 | | 18.2% | Sep 2023 | huggingface.co | Official leaderboard | |
| 8 | PaLM-2 closed weights, 340B | 16.5% | Official leaderboard | |||
| 9 | | 9.21% | Dec 2023 | Official leaderboard | ||
| 10 | | 7.56% | Jun 2023 | huggingface.co | Official leaderboard | |
| 11 | | 0.22% | Sep 2023 | Official leaderboard | ||
| 12 | mosaicml/mpt-30b-instruct open weights, 30B | 0.02% | Official leaderboard |
| # | Model | Score | Setting | Date | Link | Source |
|---|---|---|---|---|---|---|
| 1 | GPT-4 Turbo Apr 2024 | 37.7% | Apr 2024 | Official leaderboard | ||
| 2 | Claude 2 | 26.4% | Jul 2023 | Official leaderboard | ||
| 3 | GPT-3.5 Turbo Jan 2024 | 25.9% | Jan 2024 | Official leaderboard | ||
| 4 | Claude Instant | 23.9% | Aug 2023 | Official leaderboard | ||
| 5 | | 22.3% | Sep 2023 | huggingface.co | Official leaderboard | |
| 6 | | 18.2% | Sep 2023 | huggingface.co | Official leaderboard | |
| 7 | | 9.21% | Dec 2023 | Official leaderboard | ||
| 8 | | 7.56% | Jun 2023 | huggingface.co | Official leaderboard | |
| 9 | | 0.22% | Sep 2023 | Official leaderboard |
Rankings by part
We read 9 parts from the benchmark’s own leaderboard, each ranking models on one part of the test, such as SAS, SPP and EDP. GPT-4 Turbo leads 7 of them, and 2 other models lead the rest. Arrows show how far a model moved from its overall rank among the models we track.
SAS
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 100%
- 2 GPT-3.5 Turbo Jan 2024 ↑1 94.2%
- 3
Qwen-14B ↑2 70.5%
- 4 Claude 2 ↓2 44.5%
- 5 Claude Instant ↓1 44.2%
- 6
Phi-2 ↑1 19.1%
- 7
Mistral-7B-Instruct-v0.1 ↓1 14.9%
- 8
Vicuna-13B-v1.3 11.3%
- 9
Phi-1.5 0%
SPP
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 62.9%
- 2 Claude 2 37.3%
- 3 Claude Instant ↑1 26%
- 4 GPT-3.5 Turbo Jan 2024 ↓1 22.2%
- 5
Phi-2 ↑2 3.27%
- 6
Qwen-14B ↓1 1.82%
- 7
Mistral-7B-Instruct-v0.1 ↓1 1.64%
- 8
Vicuna-13B-v1.3 0%
- 9
Phi-1.5 0%
EDP
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 53.6%
- 2 GPT-3.5 Turbo Jan 2024 ↑1 31.8%
- 3
Qwen-14B ↑2 26.9%
- 4 Claude Instant 17.6%
- 5
Vicuna-13B-v1.3 ↑3 14.7%
- 6 Claude 2 ↓4 12%
- 7
Mistral-7B-Instruct-v0.1 ↓1 5.82%
- 8
Phi-2 ↓1 0.91%
- 9
Phi-1.5 0%
TSP_D
9 models we track have a score here.
- 1 Claude 2 ↑1 87.3%
- 2 Claude Instant ↑2 72.4%
- 3
Mistral-7B-Instruct-v0.1 ↑3 62.7%
- 4
Qwen-14B ↑1 41.6%
- 5 GPT-3.5 Turbo Jan 2024 ↓2 21.4%
- 6 GPT-4 Turbo Apr 2024 ↓5 14%
- 7
Vicuna-13B-v1.3 ↑1 2.91%
- 8
Phi-2 ↓1 1.45%
- 9
Phi-1.5 0%
GCP_D
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 73.3%
- 2
Mistral-7B-Instruct-v0.1 ↑4 58%
- 3
Qwen-14B ↑2 56%
- 4
Phi-2 ↑3 55.8%
- 5 GPT-3.5 Turbo Jan 2024 ↓2 52.5%
- 6 Claude 2 ↓4 52.2%
- 7 Claude Instant ↓3 50.2%
- 8
Vicuna-13B-v1.3 34.4%
- 9
Phi-1.5 0%
KSP
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 18.7%
- 2 Claude 2 0.18%
- 3
Vicuna-13B-v1.3 ↑5 0%
- 4 Claude Instant 0%
- 5
Phi-1.5 ↑4 0%
- 6
Qwen-14B ↓1 0%
- 7
Mistral-7B-Instruct-v0.1 ↓1 0%
- 8
Phi-2 ↓1 0%
- 9 GPT-3.5 Turbo Jan 2024 ↓6 0%
TSP
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 8.18%
- 2 Claude 2 2.00%
- 3 Claude Instant ↑1 2.00%
- 4 GPT-3.5 Turbo Jan 2024 ↓1 1.64%
- 5
Qwen-14B 1.27%
- 6
Phi-2 ↑1 1.09%
- 7
Vicuna-13B-v1.3 ↑1 0%
- 8
Phi-1.5 ↑1 0%
- 9
Mistral-7B-Instruct-v0.1 ↓3 0%
GCP
9 models we track have a score here.
- 1
Mistral-7B-Instruct-v0.1 ↑5 20.9%
- 2 GPT-3.5 Turbo Jan 2024 ↑1 8.36%
- 3 GPT-4 Turbo Apr 2024 ↓2 7.64%
- 4
Vicuna-13B-v1.3 ↑4 4.73%
- 5 Claude Instant ↓1 2.91%
- 6 Claude 2 ↓4 2.36%
- 7
Qwen-14B ↓2 2.36%
- 8
Phi-1.5 ↑1 2.00%
- 9
Phi-2 ↓2 1.27%
MSP
9 models we track have a score here.
- 1 GPT-4 Turbo Apr 2024 1.27%
- 2 GPT-3.5 Turbo Jan 2024 ↑1 0.55%
- 3 Claude Instant ↑1 0.18%
- 4
Vicuna-13B-v1.3 ↑4 0%
- 5 Claude 2 ↓3 0%
- 6
Phi-1.5 ↑3 0%
- 7
Qwen-14B ↓2 0%
- 8
Mistral-7B-Instruct-v0.1 ↓2 0%
- 9
Phi-2 ↓2 0%
2 models have held number one on the official leaderboard since its release in December 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official NPHardEval 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 NPHardEval 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 2 years ago, in April 2024, for GPT-4 Turbo. 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 (12 entries).
- Pass Models scored 9 models we track, plus 3 other entries on its leaderboard.
- Warning Newest score The newest came 2 years ago, in April 2024, for GPT-4 Turbo.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-4 Turbo (#172).
- Pass Headroom The best is 37.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on huggingface.co: 9 rows for models we track and 3 other entries 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.