SATA-Bench
Models select every correct option in 1,650 multiple-choice questions. Exact set match measures complete answers, with Jaccard overlap and option-count metrics as diagnostics.
Decision Index uses an adapted scoring protocol. Do not combine exact match with Jaccard scores. The maker repository does not establish a release month.
- Made by
- SATA-Bench authors
- Kind
- Language reasoning
- Level
- University
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
- o3 #84 on the capability index
- Last model added
- August 2025
- Scores refreshed
- November 21st, 2026
GPT-4.1 leads with 75.2% among models released by August 2025 GPT-4.1 leads with 75.2% among models released by August 2025
Best score per entry from the benchmark's paper, read on November 21st, 2026. A paper’s table doesn’t add newer models. 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 August 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
- 75.2 GPT-4.1
- 74.4 Grok 3
- 74.3 gpt-oss-120b
- 74.1 GPT-4
- 73.9 o3
- 72.6 Geimini 2.5 Think
- 71.1 Claude 3.5 Haiku
- 71.0 Claude 3.7
- 71.0 Claude 3.7 Think
- 70.7 Claude 3 Sonnet
- 70.7
Gemini 2.0 Flash
- 70.6 Claude 3 Haiku
- 70.2 Claude 3 Opus
- 69.9 GPT-4.1 mini
- 68.9 Nova Pro
- 67.2
Llama 3.1-405B
- 67.2 Claude 3.5 Sonnet
- 64.5
DeepSeek-R1
- 63.8 Nova Lite
- 60.7 gpt-oss-20b
- 57.2 Mistral Large V2
- 55.8
Llama 3.2 90B
- 55.8 Nova Micro
- 55.7
Qwen Plus
- 55.6
Llama 3.1-70B
- 46.6 Mistral 8B
- 43.6
Llama 3-8B
- 41.1 Bloomz 7B
- 40.0 DeepSeek R1 Distill 8B
- 37.6
Qwen2.5-14B
- 34.6 Phi3 7B
- 29.7
Phi-4-mini-reasoning
- 1 GPT-4.1 Inference based retrieval and chain of thought 75.2%
- 2 Grok 3 Inference based retrieval and chain of thought 74.4%
- 3 gpt-oss-120b Inference based retrieval and chain of thought 74.3%
- 4 GPT-4 Jun 2023, Inference based retrieval and chain of thought 74.1%
- 5 o3 Inference based retrieval and chain of thought 73.9%
- 6 Geimini 2.5 Think Inference based retrieval and chain of thought 72.6%
- 7 Claude 3.5 Haiku Inference based retrieval and chain of thought 71.1%
- 8 Claude 3.7 Inference based retrieval and chain of thought 71%
- 9 Claude 3.7 Think Inference based retrieval and chain of thought 71%
- 10 Claude 3 Sonnet Inference based retrieval and chain of thought 70.7%
- 11
Gemini 2.0 Flash Feb 2025, Inference based retrieval and chain of thought 70.7%
- 12 Claude 3 Haiku Inference based retrieval and chain of thought 70.6%
- 13 Claude 3 Opus Inference based retrieval and chain of thought 70.2%
- 14 GPT-4.1 mini Inference based retrieval and chain of thought 69.9%
- 15 Nova Pro Inference based retrieval and chain of thought 68.9%
- 16
Llama 3.1-405B Inference based retrieval and chain of thought 67.2%
- 17 Claude 3.5 Sonnet Inference based retrieval and chain of thought 67.2%
- 18
DeepSeek-R1 May 2025, Inference based retrieval and chain of thought 64.5%
- 19 Nova Lite Inference based retrieval and chain of thought 63.8%
- 20 gpt-oss-20b Inference based retrieval and chain of thought 60.7%
- 21 Mistral Large V2 Inference based retrieval and chain of thought 57.2%
- 22
Llama 3.2 90B Inference based retrieval and chain of thought 55.8%
- 23 Nova Micro Inference based retrieval and chain of thought 55.8%
- 24
Qwen Plus Apr 2025, Inference based retrieval and chain of thought 55.7%
- 25
Llama 3.1-70B Inference based retrieval and chain of thought 55.6%
- 26 Mistral 8B Probability based retrieval 46.6%
- 27
Llama 3-8B Probability based retrieval 43.6%
- 28 Bloomz 7B Probability based retrieval 41.1%
- 29 DeepSeek R1 Distill 8B Probability based retrieval 40%
- 30
Qwen2.5-14B Probability based retrieval 37.6%
- 31 Phi3 7B Probability based retrieval 34.6%
- 32
Phi-4-mini-reasoning Probability based retrieval 29.7%
Source: The SATA-Bench paper
- 75.2 GPT-4.1
- 74.4 Grok 3
- 74.3 gpt-oss-120b
- 74.1 GPT-4
- 73.9 o3
- 71.1 Claude 3.5 Haiku
- 70.7 Claude 3 Sonnet
- 70.7
Gemini 2.0 Flash
- 70.6 Claude 3 Haiku
- 70.2 Claude 3 Opus
- 69.9 GPT-4.1 mini
- 67.2
Llama 3.1-405B
- 67.2 Claude 3.5 Sonnet
- 64.5
DeepSeek-R1
- 60.7 gpt-oss-20b
- 55.8
Llama 3.2 90B
- 55.7
Qwen Plus
- 55.6
Llama 3.1-70B
- 43.6
Llama 3-8B
- 37.6
Qwen2.5-14B
- 29.7
Phi-4-mini-reasoning
- 1 GPT-4.1 Inference based retrieval and chain of thought 75.2%
- 2 Grok 3 Inference based retrieval and chain of thought 74.4%
- 3 gpt-oss-120b Inference based retrieval and chain of thought 74.3%
- 4 GPT-4 Jun 2023, Inference based retrieval and chain of thought 74.1%
- 5 o3 Inference based retrieval and chain of thought 73.9%
- 6 Claude 3.5 Haiku Inference based retrieval and chain of thought 71.1%
- 7 Claude 3 Sonnet Inference based retrieval and chain of thought 70.7%
- 8
Gemini 2.0 Flash Feb 2025, Inference based retrieval and chain of thought 70.7%
- 9 Claude 3 Haiku Inference based retrieval and chain of thought 70.6%
- 10 Claude 3 Opus Inference based retrieval and chain of thought 70.2%
- 11 GPT-4.1 mini Inference based retrieval and chain of thought 69.9%
- 12
Llama 3.1-405B Inference based retrieval and chain of thought 67.2%
- 13 Claude 3.5 Sonnet Inference based retrieval and chain of thought 67.2%
- 14
DeepSeek-R1 May 2025, Inference based retrieval and chain of thought 64.5%
- 15 gpt-oss-20b Inference based retrieval and chain of thought 60.7%
- 16
Llama 3.2 90B Inference based retrieval and chain of thought 55.8%
- 17
Qwen Plus Apr 2025, Inference based retrieval and chain of thought 55.7%
- 18
Llama 3.1-70B Inference based retrieval and chain of thought 55.6%
- 19
Llama 3-8B Probability based retrieval 43.6%
- 20
Qwen2.5-14B Probability based retrieval 37.6%
- 21
Phi-4-mini-reasoning Probability based retrieval 29.7%
Source: The SATA-Bench paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-4.1 | 75.2% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 2 | Grok 3 | 74.4% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 3 | gpt-oss-120b | 74.3% | Inference based retrieval and chain of thought | Aug 2025 | Official leaderboard |
| 4 | GPT-4 Jun 2023 | 74.1% | Inference based retrieval and chain of thought | Jun 2023 | Official leaderboard |
| 5 | o3 | 73.9% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 6 | Geimini 2.5 Think | 72.6% | Inference based retrieval and chain of thought | Official leaderboard | |
| 7 | Claude 3.5 Haiku | 71.1% | Inference based retrieval and chain of thought | Oct 2024 | Official leaderboard |
| 8 | Claude 3.7 | 71% | Inference based retrieval and chain of thought | Official leaderboard | |
| 9 | Claude 3.7 Think | 71% | Inference based retrieval and chain of thought | Official leaderboard | |
| 10 | Claude 3 Sonnet | 70.7% | Inference based retrieval and chain of thought | Feb 2024 | Official leaderboard |
| 11 | | 70.7% | Inference based retrieval and chain of thought | Feb 2025 | Official leaderboard |
| 12 | Claude 3 Haiku | 70.6% | Inference based retrieval and chain of thought | Mar 2024 | Official leaderboard |
| 13 | Claude 3 Opus | 70.2% | Inference based retrieval and chain of thought | Feb 2024 | Official leaderboard |
| 14 | GPT-4.1 mini | 69.9% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 15 | Nova Pro | 68.9% | Inference based retrieval and chain of thought | Official leaderboard | |
| 16 | | 67.2% | Inference based retrieval and chain of thought | Jul 2024 | Official leaderboard |
| 17 | Claude 3.5 Sonnet | 67.2% | Inference based retrieval and chain of thought | Jun 2024 | Official leaderboard |
| 18 | | 64.5% | Inference based retrieval and chain of thought | May 2025 | Official leaderboard |
| 19 | Nova Lite | 63.8% | Inference based retrieval and chain of thought | Official leaderboard | |
| 20 | gpt-oss-20b | 60.7% | Inference based retrieval and chain of thought | Aug 2025 | Official leaderboard |
| 21 | Mistral Large V2 | 57.2% | Inference based retrieval and chain of thought | Official leaderboard | |
| 22 | | 55.8% | Inference based retrieval and chain of thought | Sep 2024 | Official leaderboard |
| 23 | Nova Micro | 55.8% | Inference based retrieval and chain of thought | Official leaderboard | |
| 24 | | 55.7% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 25 | | 55.6% | Inference based retrieval and chain of thought | Jul 2024 | Official leaderboard |
| 26 | Mistral 8B | 46.6% | Probability based retrieval | Official leaderboard | |
| 27 | | 43.6% | Probability based retrieval | Apr 2024 | Official leaderboard |
| 28 | Bloomz 7B | 41.1% | Probability based retrieval | Official leaderboard | |
| 29 | DeepSeek R1 Distill 8B | 40% | Probability based retrieval | Official leaderboard | |
| 30 | | 37.6% | Probability based retrieval | Sep 2024 | Official leaderboard |
| 31 | Phi3 7B | 34.6% | Probability based retrieval | Official leaderboard | |
| 32 | | 29.7% | Probability based retrieval | Apr 2025 | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | GPT-4.1 | 75.2% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 2 | Grok 3 | 74.4% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 3 | gpt-oss-120b | 74.3% | Inference based retrieval and chain of thought | Aug 2025 | Official leaderboard |
| 4 | GPT-4 Jun 2023 | 74.1% | Inference based retrieval and chain of thought | Jun 2023 | Official leaderboard |
| 5 | o3 | 73.9% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 6 | Claude 3.5 Haiku | 71.1% | Inference based retrieval and chain of thought | Oct 2024 | Official leaderboard |
| 7 | Claude 3 Sonnet | 70.7% | Inference based retrieval and chain of thought | Feb 2024 | Official leaderboard |
| 8 | | 70.7% | Inference based retrieval and chain of thought | Feb 2025 | Official leaderboard |
| 9 | Claude 3 Haiku | 70.6% | Inference based retrieval and chain of thought | Mar 2024 | Official leaderboard |
| 10 | Claude 3 Opus | 70.2% | Inference based retrieval and chain of thought | Feb 2024 | Official leaderboard |
| 11 | GPT-4.1 mini | 69.9% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 12 | | 67.2% | Inference based retrieval and chain of thought | Jul 2024 | Official leaderboard |
| 13 | Claude 3.5 Sonnet | 67.2% | Inference based retrieval and chain of thought | Jun 2024 | Official leaderboard |
| 14 | | 64.5% | Inference based retrieval and chain of thought | May 2025 | Official leaderboard |
| 15 | gpt-oss-20b | 60.7% | Inference based retrieval and chain of thought | Aug 2025 | Official leaderboard |
| 16 | | 55.8% | Inference based retrieval and chain of thought | Sep 2024 | Official leaderboard |
| 17 | | 55.7% | Inference based retrieval and chain of thought | Apr 2025 | Official leaderboard |
| 18 | | 55.6% | Inference based retrieval and chain of thought | Jul 2024 | Official leaderboard |
| 19 | | 43.6% | Probability based retrieval | Apr 2024 | Official leaderboard |
| 20 | | 37.6% | Probability based retrieval | Sep 2024 | Official leaderboard |
| 21 | | 29.7% | Probability based retrieval | Apr 2025 | Official leaderboard |
3 models have held number one in the benchmark’s paper since 2023
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The SATA-Bench paper
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official SATA-Bench 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 15 months ago, in August 2025, for gpt-oss-120b. 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 benchmark’s paper (32 entries).
- Pass Models scored 21 models we track, plus 11 other entries on its leaderboard.
- Warning Newest score The newest came 15 months ago, in August 2025, for gpt-oss-120b.
- Note Frontier models None of the top 10 has a score. The strongest tested is o3 (#84).
- Pass Headroom The best is 75.2%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 21 rows for models we track and 11 other entries at the last read.
- Pass Last read November 21st, 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.