legal benchmark
MMLU-Pro Leaderboard
A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks. Features over 12,000 curated questions across 14 domains and causes a 16-33% accuracy drop compared to original MMLU.
Updated Aug 17, 2026
Models100
Model coverage134
MetricScore
EvidenceB
MMLU-Pro Ranking
Higher score ranks better on this benchmark.
Sort by
Rank | Model | Score | Percentile | Participants | Evidence | Evaluated |
|---|
| Rank01 | ModelSA | Score90.3% | Percentile100.0% | Participants134 | EvidenceC | Evaluated |
| Rank02 | ModelAC | Score89.6% | Percentile99.3% | Participants134 | EvidenceC | Evaluated |
| Rank03 | ModelAC | Score88.5% | Percentile98.5% | Participants134 | EvidenceC | Evaluated |
| Rank04 | ModelAC | Score88.5% | Percentile97.7% | Participants134 | EvidenceC | Evaluated |
| Rank05 | ModelMI | Score88.0% | Percentile97.0% | Participants134 | EvidenceC | Evaluated |
| Rank06 | ModelAC | Score87.8% | Percentile96.2% | Participants134 | EvidenceC | Evaluated |
| Rank07 | ModelDE | Score87.5% | Percentile95.5% | Participants134 | EvidenceC | Evaluated |
| Rank08 | ModelMA | Score87.1% | Percentile94.7% | Participants134 | EvidenceC | Evaluated |
| Rank09 | ModelBA | Score87.0% | Percentile94.0% | Participants134 | EvidenceC | Evaluated |
| Rank10 | ModelNV | Score86.8% | Percentile93.2% | Participants134 | EvidenceC | Evaluated |
| Rank11 | ModelAC | Score86.7% | Percentile92.5% | Participants134 | EvidenceC | Evaluated |
| Rank12 | ModelDE | Score86.4% | Percentile91.7% | Participants134 | EvidenceC | Evaluated |
| Rank13 | ModelUP | Score86.3% | Percentile91.0% | Participants134 | EvidenceC | Evaluated |
| Rank14 | ModelDE | Score86.2% | Percentile90.2% | Participants134 | EvidenceC | Evaluated |
| Rank15 | ModelAC | Score86.2% | Percentile89.5% | Participants134 | EvidenceC | Evaluated |
| Rank16 | ModelAC | Score86.1% | Percentile88.7% | Participants134 | EvidenceC | Evaluated |
| Rank17 | ModelAC | Score85.3% | Percentile88.0% | Participants134 | EvidenceC | Evaluated |
| Rank18 | ModelGO | Score85.2% | Percentile87.2% | Participants134 | EvidenceC | Evaluated |
| Rank19 | ModelAC | Score85.2% | Percentile86.5% | Participants134 | EvidenceC | Evaluated |
| Rank20 | ModelDE | Score85.0% | Percentile85.7% | Participants134 | EvidenceC | Evaluated |
| Rank21 | ModelDE | Score85.0% | Percentile85.0% | Participants134 | EvidenceC | Evaluated |
| Rank22 | ModelDE | Score85.0% | Percentile84.2% | Participants134 | EvidenceC | Evaluated |
| Rank23 | ModelDE | Score85.0% | Percentile83.5% | Participants134 | EvidenceC | Evaluated |
| Rank24 | ModelMI | Score85.0% | Percentile82.7% | Participants134 | EvidenceC | Evaluated |
| Rank25 | ModelXI | Score84.9% | Percentile82.0% | Participants134 | EvidenceC | Evaluated |
| Rank26 | ModelZA | Score84.6% | Percentile81.2% | Participants134 | EvidenceC | Evaluated |
| Rank27 | ModelMA | Score84.6% | Percentile80.5% | Participants134 | EvidenceC | Evaluated |
| Rank28 | ModelAC | Score84.4% | Percentile79.7% | Participants134 | EvidenceC | Evaluated |
| Rank29 | ModelZA | Score84.3% | Percentile79.0% | Participants134 | EvidenceC | Evaluated |
| Rank30 | ModelLA | Score83.8% | Percentile78.2% | Participants134 | EvidenceC | Evaluated |
| Rank31 | ModelAC | Score83.8% | Percentile77.4% | Participants134 | EvidenceC | Evaluated |
| Rank32 | ModelNV | Score83.7% | Percentile76.7% | Participants134 | EvidenceC | Evaluated |
| Rank33 | ModelDE | Score83.7% | Percentile75.9% | Participants134 | EvidenceC | Evaluated |
| Rank34 | ModelAC | Score83.0% | Percentile75.2% | Participants134 | EvidenceC | Evaluated |
| Rank35 | ModelAC | Score82.7% | Percentile74.4% | Participants134 | EvidenceC | Evaluated |
| Rank36 | ModelME | Score82.7% | Percentile73.7% | Participants134 | EvidenceC | Evaluated |
| Rank37 | ModelGO | Score82.6% | Percentile72.9% | Participants134 | EvidenceC | Evaluated |
| Rank38 | ModelME | Score82.6% | Percentile72.2% | Participants134 | EvidenceC | Evaluated |
| Rank39 | ModelMA | Score82.5% | Percentile71.4% | Participants134 | EvidenceC | Evaluated |
| Rank40 | ModelAC | Score82.5% | Percentile70.7% | Participants134 | EvidenceC | Evaluated |
| Rank41 | ModelAC | Score82.1% | Percentile69.9% | Participants134 | EvidenceC | Evaluated |
| Rank42 | ModelMI | Score82.0% | Percentile69.2% | Participants134 | EvidenceC | Evaluated |
| Rank43 | ModelNV | Score81.9% | Percentile68.4% | Participants134 | EvidenceC | Evaluated |
| Rank44 | ModelAC | Score81.8% | Percentile67.7% | Participants134 | EvidenceC | Evaluated |
| Rank45 | ModelSA | Score81.7% | Percentile66.9% | Participants134 | EvidenceC | Evaluated |
| Rank46 | ModelAM Nova 2 ProAmazon | Score81.6% | Percentile66.2% | Participants134 | EvidenceC | Evaluated |
| Rank47 | ModelZA | Score81.4% | Percentile65.4% | Participants134 | EvidenceC | Evaluated |
| Rank48 | ModelDE | Score81.2% | Percentile64.7% | Participants134 | EvidenceC | Evaluated |
| Rank49 | ModelMA | Score81.1% | Percentile63.9% | Participants134 | EvidenceC | Evaluated |
| Rank50 | ModelMA | Score81.1% | Percentile63.2% | Participants134 | EvidenceC | Evaluated |
| Rank51 | ModelMI | Score81.1% | Percentile62.4% | Participants134 | EvidenceC | Evaluated |
| Rank52 | ModelAMNova 2 LiteAmazon | Score80.9% | Percentile61.6% | Participants134 | EvidenceC | Evaluated |
| Rank53 | ModelOP | Score80.7% | Percentile60.9% | Participants134 | EvidenceC | Evaluated |
| Rank54 | ModelAMNova 2 OmniAmazon | Score80.7% | Percentile60.1% | Participants134 | EvidenceC | Evaluated |
| Rank55 | ModelMI | Score80.6% | Percentile59.4% | Participants134 | EvidenceC | Evaluated |
| Rank56 | ModelAC | Score80.6% | Percentile58.6% | Participants134 | EvidenceC | Evaluated |
| Rank57 | ModelME | Score80.5% | Percentile57.9% | Participants134 | EvidenceC | Evaluated |
| Rank58 | ModelAC | Score80.5% | Percentile57.1% | Participants134 | EvidenceC | Evaluated |
| Rank59 | ModelSA | Score80.0% | Percentile56.4% | Participants134 | EvidenceC | Evaluated |
| Rank60 | ModelAC | Score79.1% | Percentile55.6% | Participants134 | EvidenceC | Evaluated |
| Rank61 | ModelAC | Score78.6% | Percentile54.9% | Participants134 | EvidenceC | Evaluated |
| Rank62 | ModelNV | Score78.3% | Percentile54.1% | Participants134 | EvidenceC | Evaluated |
| Rank63 | ModelME | Score78.3% | Percentile53.4% | Participants134 | EvidenceC | Evaluated |
| Rank64 | ModelMA | Score78.0% | Percentile52.6% | Participants134 | EvidenceC | Evaluated |
| Rank65 | ModelAC | Score77.8% | Percentile51.9% | Participants134 | EvidenceC | Evaluated |
| Rank66 | ModelAN | Score77.6% | Percentile51.1% | Participants134 | EvidenceC | Evaluated |
| Rank67 | ModelGO | Score77.6% | Percentile50.4% | Participants134 | EvidenceC | Evaluated |
| Rank68 | ModelAC | Score77.3% | Percentile49.6% | Participants134 | EvidenceC | Evaluated |
| Rank69 | ModelGO | Score77.2% | Percentile48.9% | Participants134 | EvidenceC | Evaluated |
| Rank70 | ModelGO | Score76.4% | Percentile48.1% | Participants134 | EvidenceC | Evaluated |
| Rank71 | ModelAN | Score76.1% | Percentile47.4% | Participants134 | EvidenceC | Evaluated |
| Rank72 | ModelMI | Score76.0% | Percentile46.6% | Participants134 | EvidenceC | Evaluated |
| Rank73 | ModelDE | Score75.9% | Percentile45.9% | Participants134 | EvidenceC | Evaluated |
| Rank74 | ModelGO | Score75.8% | Percentile45.1% | Participants134 | EvidenceC | Evaluated |
| Rank75 | ModelXA | Score75.5% | Percentile44.4% | Participants134 | EvidenceC | Evaluated |
| Rank76 | ModelOP | Score74.7% | Percentile43.6% | Participants134 | EvidenceC | Evaluated |
| Rank77 | ModelME | Score74.3% | Percentile42.9% | Participants134 | EvidenceC | Evaluated |
| Rank78 | ModelMI | Score74.3% | Percentile42.1% | Participants134 | EvidenceC | Evaluated |
| Rank79 | ModelAC | Score73.6% | Percentile41.4% | Participants134 | EvidenceC | Evaluated |
| Rank80 | ModelME | Score73.3% | Percentile40.6% | Participants134 | EvidenceC | Evaluated |
| Rank81 | ModelOP | Score72.6% | Percentile39.9% | Participants134 | EvidenceC | Evaluated |
| Rank82 | ModelXA | Score72.0% | Percentile39.1% | Participants134 | EvidenceC | Evaluated |
| Rank83 | ModelGO | Score71.6% | Percentile38.4% | Participants134 | EvidenceC | Evaluated |
| Rank84 | ModelAC | Score71.6% | Percentile37.6% | Participants134 | EvidenceC | Evaluated |
| Rank85 | ModelAC | Score71.1% | Percentile36.8% | Participants134 | EvidenceC | Evaluated |
| Rank86 | ModelMI | Score70.4% | Percentile36.1% | Participants134 | EvidenceC | Evaluated |
| Rank87 | ModelGO | Score69.4% | Percentile35.3% | Participants134 | EvidenceC | Evaluated |
| Rank88 | ModelMA | Score69.2% | Percentile34.6% | Participants134 | EvidenceC | Evaluated |
| Rank89 | ModelMA | Score69.1% | Percentile33.8% | Participants134 | EvidenceC | Evaluated |
| Rank90 | ModelAC | Score69.0% | Percentile33.1% | Participants134 | EvidenceC | Evaluated |
| Rank91 | ModelME | Score68.9% | Percentile32.3% | Participants134 | EvidenceC | Evaluated |
| Rank92 | ModelAC | Score68.8% | Percentile31.6% | Participants134 | EvidenceC | Evaluated |
| Rank93 | ModelAN | Score68.5% | Percentile30.8% | Participants134 | EvidenceC | Evaluated |
| Rank94 | ModelXI | Score68.5% | Percentile30.1% | Participants134 | EvidenceC | Evaluated |
| Rank95 | ModelAC | Score68.2% | Percentile29.3% | Participants134 | EvidenceC | Evaluated |
| Rank96 | ModelGO | Score67.5% | Percentile28.6% | Participants134 | EvidenceC | Evaluated |
| Rank97 | ModelGO | Score67.3% | Percentile27.8% | Participants134 | EvidenceC | Evaluated |
| Rank98 | ModelAC | Score67.1% | Percentile27.1% | Participants134 | EvidenceC | Evaluated |
| Rank99 | ModelOP | Score67.0% | Percentile26.3% | Participants134 | EvidenceC | Evaluated |
| Rank100 | ModelMA | Score66.8% | Percentile25.6% | Participants134 | EvidenceC | Evaluated |
MMLU-Pro Highlights
The leading models and scores on this benchmark.
MMLU-Pro Score Distribution
A closer view of the leading scores on this benchmark.
The Top AI Models for MMLU-Pro
The first five results on this benchmark, with official price and output speed added where the model identity can be matched.
Ranking basisThis mmlu-pro AI model leaderboard uses descending score in the benchmark's original unit. The leaderboard ranking keeps matched price and speed data separate from benchmark evidence.
- 01SASakana NamazuSakana AI
- Score
- 90.3%
Strengths
- Ranks #1 of 134 compared models
- 100th percentile on this benchmark
- C evidence result
Considerations
- This result measures MMLU-Pro, not total model capability
- 02ACQwen3.7 MaxAlibaba Cloud / Qwen Team
- Score
- 89.6%
- Price
- $2.5 input / $7.5 output per 1M tokens
- Speed
- Up to 5.8 tok/s via Together
Strengths
- Ranks #2 of 134 compared models
- 99th percentile on this benchmark
- C evidence result
Considerations
- This result measures MMLU-Pro, not total model capability
- 03ACQwen3.6 PlusAlibaba Cloud / Qwen Team
- Score
- 88.5%
- Price
- $0.50 input / $3.0 output per 1M tokens
- Speed
- Up to 16 tok/s via Together
Strengths
- Ranks #3 of 134 compared models
- 99th percentile on this benchmark
- C evidence result
Considerations
- This result measures MMLU-Pro, not total model capability
- 04ACQwen3.7-PlusAlibaba Cloud / Qwen Team
- Score
- 88.5%
- Price
- $0.50 input / $3.0 output per 1M tokens
Strengths
- Ranks #4 of 134 compared models
- 98th percentile on this benchmark
- C evidence result
Considerations
- This result measures MMLU-Pro, not total model capability
- 05MIMiniMax M2.1MiniMax
- Score
- 88.0%
- Price
- $0.30 input / $1.2 output per 1M tokens
- Speed
- Up to 100 tok/s via MiniMax
Strengths
- Ranks #5 of 134 compared models
- 97th percentile on this benchmark
- C evidence result
Considerations
- This result measures MMLU-Pro, not total model capability
Selection summary
Best AI Models for MMLU-Pro
Sakana Namazu currently leads MMLU-Pro with 90.3%. It is the top model on this specific benchmark, while the best LLM for the broader task should also be checked against other benchmarks, price and runtime.
Use this leaderboard with the supporting benchmark results and coverage details above. A leaderboard position summarizes the selected ranking signal; it does not replace workload-specific testing.
What is MMLU-Pro?
What MMLU-Pro measures and how its scores work.
A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks. Features over 12,000 curated questions across 14 domains and causes a 16-33% accuracy drop compared to original MMLU.
Scores are shown in ratio. This benchmark is not independently verified and has an evidence level of B.
- Family
- MMLU-Pro
- Modality
- text
- Primary category
- legal
- Score direction
- higher
- LLMBoard eligible
- Yes
- Evaluation key
- mmlu-pro|llm-stats-current
Benchmark scores retain their original unit. Overall score eligibility is shown separately.
FAQ
Common questions about MMLU-Pro.
Which model scores highest on MMLU-Pro?
Sakana Namazu is currently ranked first with 90.3%.
What does MMLU-Pro measure?
A more robust and challenging multi-task language understanding benchmark that extends MMLU by expanding multiple-choice options from 4 to 10, eliminating trivial questions, and focusing on reasoning-intensive tasks. Features over 12,000 curated questions across 14 domains and causes a 16-33% accuracy drop compared to original MMLU.
Is a higher score better?
Yes. Higher values rank better for this benchmark.
How many models are compared?
100 model results are currently shown.
Does this benchmark affect the overall score?
Yes. This benchmark can contribute to the current LLMBoard capability score.