South Asian Music Understanding
Models answer questions about Hindustani music theory and Bengali musical traditions. They also choose a continuation for a melody written in ABC notation.
The authors publish accuracy for 33 models on three tasks. Their 198 continuation rows contain 150 distinct combinations of question, answer choices, and correct answer.
- Made by
- University of Central Florida and Auburn University
- Released
- June 2026
- Kind
- Expert-written questions
- Level
- University
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Subject
- Arts
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- Strongest model tested
-
Phi-4 #160 on the capability index
- Last model added
- March 2025
- Scores refreshed
- November 20th, 2026
Gemini 2.5 Pro (snapshot unspecified) tops the leaderboard with 88.5% Qwen2.5-32B leads with 48.7% among models released by March 2025
Best score per entry from the benchmark's paper, read on November 20th, 2026. A paper’s table doesn’t add newer models. The authors report no overall score, so this is the plain mean of accuracy across the three music understanding tasks. 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. The best model we track is Qwen2.5-32B, at 48.7%. This view shows only models we track, each at its best.
The newest model with a score here came out in March 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
- 88.5 Gemini 2.5 Pro (snapshot unspecified)
- 65.3 GPT-4o (snapshot unspecified)
- 57.1 GPT-3.5 (snapshot unspecified)
- 48.7
Qwen2.5-32B
- 48.2
Qwen2.5-14B
- 46.4
Phi-4
- 45.9
QwQ-32B-Preview
- 45.6
DeepSeek-R1-Distill-Qwen-32B
- 43.8 Cogito-v1-preview-llama-8B
- 42.7
Qwen2.5-7B
- 42.2
DeepSeek-R1-Distill-Qwen-14B
- 39.1
DeepSeek-R1-Distill-Qwen-7B
- 37.3
Falcon3-10B-Instruct
- 34.9
Qwen2.5-3B
- 34.7
Llama 3.1-8B
- 34.5
Falcon3-7B-Instruct
- 32.2
Phi-3.5-mini
- 31.8 DeepSeek-R1-Distill-Llama-8B
- 29.8
Phi-4 Mini
- 28.7
Gemma 2 2B IT
- 28.1
Qwen2.5-1.5B
- 27.6 OLMo-2-1124-13B-Instruct
- 26.8
Llama 3.2 3B
- 26.3
Falcon3-3B-Instruct
- 25.8 Cogito-v1-preview-llama-3B
- 25.5 OLMo-7B-0724-Instruct
- 25.2
Qwen2.5-0.5B
- 24.9
DeepSeek-R1-Distill-Qwen-1.5B
- 24.8 OLMo-2-1124-7B-Instruct
- 24.3
Gemma 3 1B
- 23.7
Mistral 7B v0.3
- 20.1
Llama 3.2 1B
- 8.70 ChatMusician
- 1 Gemini 2.5 Pro (snapshot unspecified) 88.5%
- 2 GPT-4o (snapshot unspecified) 65.3%
- 3 GPT-3.5 (snapshot unspecified) 57.1%
- 4
Qwen2.5-32B 48.7%
- 5
Qwen2.5-14B 48.2%
- 6
Phi-4 46.4%
- 7
QwQ-32B-Preview 45.9%
- 8
DeepSeek-R1-Distill-Qwen-32B 45.6%
- 9 Cogito-v1-preview-llama-8B 43.8%
- 10
Qwen2.5-7B 42.7%
- 11
DeepSeek-R1-Distill-Qwen-14B 42.2%
- 12
DeepSeek-R1-Distill-Qwen-7B 39.1%
- 13
Falcon3-10B-Instruct 37.3%
- 14
Qwen2.5-3B 34.9%
- 15
Llama 3.1-8B 34.7%
- 16
Falcon3-7B-Instruct 34.5%
- 17
Phi-3.5-mini 32.2%
- 18 DeepSeek-R1-Distill-Llama-8B 31.8%
- 19
Phi-4 Mini 29.8%
- 20
Gemma 2 2B IT 28.7%
- 21
Qwen2.5-1.5B 28.1%
- 22 OLMo-2-1124-13B-Instruct 27.6%
- 23
Llama 3.2 3B 26.8%
- 24
Falcon3-3B-Instruct 26.3%
- 25 Cogito-v1-preview-llama-3B 25.8%
- 26 OLMo-7B-0724-Instruct 25.5%
- 27
Qwen2.5-0.5B 25.2%
- 28
DeepSeek-R1-Distill-Qwen-1.5B 24.9%
- 29 OLMo-2-1124-7B-Instruct 24.8%
- 30
Gemma 3 1B 24.3%
- 31
Mistral 7B v0.3 23.7%
- 32
Llama 3.2 1B 20.1%
- 33 ChatMusician 8.70%
Source: The South Asian Music Understanding paper
- 48.7
Qwen2.5-32B
- 48.2
Qwen2.5-14B
- 46.4
Phi-4
- 45.9
QwQ-32B-Preview
- 45.6
DeepSeek-R1-Distill-Qwen-32B
- 42.7
Qwen2.5-7B
- 42.2
DeepSeek-R1-Distill-Qwen-14B
- 39.1
DeepSeek-R1-Distill-Qwen-7B
- 37.3
Falcon3-10B-Instruct
- 34.9
Qwen2.5-3B
- 34.7
Llama 3.1-8B
- 34.5
Falcon3-7B-Instruct
- 32.2
Phi-3.5-mini
- 29.8
Phi-4 Mini
- 28.7
Gemma 2 2B IT
- 28.1
Qwen2.5-1.5B
- 26.8
Llama 3.2 3B
- 26.3
Falcon3-3B-Instruct
- 25.2
Qwen2.5-0.5B
- 24.9
DeepSeek-R1-Distill-Qwen-1.5B
- 24.3
Gemma 3 1B
- 23.7
Mistral 7B v0.3
- 20.1
Llama 3.2 1B
- 1
Qwen2.5-32B 48.7%
- 2
Qwen2.5-14B 48.2%
- 3
Phi-4 46.4%
- 4
QwQ-32B-Preview 45.9%
- 5
DeepSeek-R1-Distill-Qwen-32B 45.6%
- 6
Qwen2.5-7B 42.7%
- 7
DeepSeek-R1-Distill-Qwen-14B 42.2%
- 8
DeepSeek-R1-Distill-Qwen-7B 39.1%
- 9
Falcon3-10B-Instruct 37.3%
- 10
Qwen2.5-3B 34.9%
- 11
Llama 3.1-8B 34.7%
- 12
Falcon3-7B-Instruct 34.5%
- 13
Phi-3.5-mini 32.2%
- 14
Phi-4 Mini 29.8%
- 15
Gemma 2 2B IT 28.7%
- 16
Qwen2.5-1.5B 28.1%
- 17
Llama 3.2 3B 26.8%
- 18
Falcon3-3B-Instruct 26.3%
- 19
Qwen2.5-0.5B 25.2%
- 20
DeepSeek-R1-Distill-Qwen-1.5B 24.9%
- 21
Gemma 3 1B 24.3%
- 22
Mistral 7B v0.3 23.7%
- 23
Llama 3.2 1B 20.1%
Source: The South Asian Music Understanding paper
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | Gemini 2.5 Pro (snapshot unspecified) | 88.5% | Official leaderboard | ||
| 2 | GPT-4o (snapshot unspecified) | 65.3% | Official leaderboard | ||
| 3 | GPT-3.5 (snapshot unspecified) | 57.1% | Official leaderboard | ||
| 4 | | 48.7% | Sep 2024 | Official leaderboard | |
| 5 | | 48.2% | Sep 2024 | Official leaderboard | |
| 6 | | 46.4% | Dec 2024 | Official leaderboard | |
| 7 | | 45.9% | Nov 2024 | Official leaderboard | |
| 8 | | 45.6% | Jan 2025 | Official leaderboard | |
| 9 | Cogito-v1-preview-llama-8B | 43.8% | Official leaderboard | ||
| 10 | | 42.7% | Sep 2024 | Official leaderboard | |
| 11 | | 42.2% | Jan 2025 | Official leaderboard | |
| 12 | | 39.1% | Jan 2025 | Official leaderboard | |
| 13 | | 37.3% | Dec 2024 | Official leaderboard | |
| 14 | | 34.9% | Sep 2024 | Official leaderboard | |
| 15 | | 34.7% | Jul 2024 | Official leaderboard | |
| 16 | | 34.5% | Dec 2024 | Official leaderboard | |
| 17 | | 32.2% | Aug 2024 | Official leaderboard | |
| 18 | DeepSeek-R1-Distill-Llama-8B | 31.8% | Official leaderboard | ||
| 19 | | 29.8% | Feb 2025 | Official leaderboard | |
| 20 | | 28.7% | Jul 2024 | Official leaderboard | |
| 21 | | 28.1% | Sep 2024 | Official leaderboard | |
| 22 | OLMo-2-1124-13B-Instruct | 27.6% | Official leaderboard | ||
| 23 | | 26.8% | Sep 2024 | Official leaderboard | |
| 24 | | 26.3% | Dec 2024 | Official leaderboard | |
| 25 | Cogito-v1-preview-llama-3B | 25.8% | Official leaderboard | ||
| 26 | OLMo-7B-0724-Instruct | 25.5% | Official leaderboard | ||
| 27 | | 25.2% | Sep 2024 | Official leaderboard | |
| 28 | | 24.9% | Jan 2025 | Official leaderboard | |
| 29 | OLMo-2-1124-7B-Instruct | 24.8% | Official leaderboard | ||
| 30 | | 24.3% | Mar 2025 | Official leaderboard | |
| 31 | | 23.7% | Sep 2023 | Official leaderboard | |
| 32 | | 20.1% | Sep 2024 | Official leaderboard | |
| 33 | ChatMusician | 8.70% | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 48.7% | Sep 2024 | Official leaderboard | |
| 2 | | 48.2% | Sep 2024 | Official leaderboard | |
| 3 | | 46.4% | Dec 2024 | Official leaderboard | |
| 4 | | 45.9% | Nov 2024 | Official leaderboard | |
| 5 | | 45.6% | Jan 2025 | Official leaderboard | |
| 6 | | 42.7% | Sep 2024 | Official leaderboard | |
| 7 | | 42.2% | Jan 2025 | Official leaderboard | |
| 8 | | 39.1% | Jan 2025 | Official leaderboard | |
| 9 | | 37.3% | Dec 2024 | Official leaderboard | |
| 10 | | 34.9% | Sep 2024 | Official leaderboard | |
| 11 | | 34.7% | Jul 2024 | Official leaderboard | |
| 12 | | 34.5% | Dec 2024 | Official leaderboard | |
| 13 | | 32.2% | Aug 2024 | Official leaderboard | |
| 14 | | 29.8% | Feb 2025 | Official leaderboard | |
| 15 | | 28.7% | Jul 2024 | Official leaderboard | |
| 16 | | 28.1% | Sep 2024 | Official leaderboard | |
| 17 | | 26.8% | Sep 2024 | Official leaderboard | |
| 18 | | 26.3% | Dec 2024 | Official leaderboard | |
| 19 | | 25.2% | Sep 2024 | Official leaderboard | |
| 20 | | 24.9% | Jan 2025 | Official leaderboard | |
| 21 | | 24.3% | Mar 2025 | Official leaderboard | |
| 22 | | 23.7% | Sep 2023 | Official leaderboard | |
| 23 | | 20.1% | Sep 2024 | Official leaderboard |
Rankings by part
We read 3 parts from the benchmark’s paper: Music Theory Understanding, Music General Knowledge and Music Continuation. Each ranks models on one part of the test. QwQ-32B-Preview leads 1 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.
Music Theory Understanding
23 models we track have a score here.
- 1
QwQ-32B-Preview ↑3 49.1%
- 2
DeepSeek-R1-Distill-Qwen-14B ↑5 39.3%
- 3
DeepSeek-R1-Distill-Qwen-32B ↑2 39.3%
- 4
Qwen2.5-14B ↓2 36.8%
- 5
Qwen2.5-32B ↓4 36.2%
- 6
DeepSeek-R1-Distill-Qwen-7B ↑2 36.2%
- 7
Qwen2.5-7B ↓1 35.6%
- 8
Llama 3.1-8B ↑3 35%
- 9
Qwen2.5-1.5B ↑7 31.9%
- 10
Phi-4 ↓7 31.9%
Music General Knowledge
23 models we track have a score here.
- 1
Qwen2.5-14B ↑1 46.9%
- 2
QwQ-32B-Preview ↑2 45.5%
- 3
Phi-4 45.5%
- 4
Qwen2.5-32B ↓3 40.6%
- 5
DeepSeek-R1-Distill-Qwen-7B ↑3 40%
- 6
DeepSeek-R1-Distill-Qwen-32B ↓1 37.8%
- 7
Qwen2.5-7B ↓1 35.7%
- 8
DeepSeek-R1-Distill-Qwen-14B ↓1 35%
- 9
Falcon3-10B-Instruct 32.9%
- 10
Falcon3-7B-Instruct ↑2 32.9%
Music Continuation
23 models we track have a score here.
- 1
Qwen2.5-32B 69.3%
- 2
Phi-4 ↑1 61.8%
- 3
Qwen2.5-14B ↓1 60.8%
- 4
DeepSeek-R1-Distill-Qwen-32B ↑1 59.8%
- 5
Qwen2.5-7B ↑1 56.8%
- 6
Qwen2.5-3B ↑4 52.8%
- 7
DeepSeek-R1-Distill-Qwen-14B 52.3%
- 8
Falcon3-10B-Instruct ↑1 51.5%
- 9
QwQ-32B-Preview ↓5 43.2%
- 10
Phi-3.5-mini ↑3 41.7%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: The South Asian Music Understanding 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 South Asian Music Understanding 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 20 months ago, in March 2025, for Gemma 3 1B. 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 (33 entries).
- Pass Models scored 23 models we track, plus 10 other entries on its leaderboard.
- Warning Newest score The newest came 20 months ago, in March 2025, for Gemma 3 1B.
- Note Frontier models None of the top 10 has a score. The strongest tested is Phi-4 (#160).
- Pass Headroom The best is 48.7%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 23 rows for models we track and 10 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.