MusICA-MetaBench
Models answer the same music questions from audio, printed scores, and ABC notation. The authors publish fixed tests for two collections of choral music.
The authors compare eight systems on each 300-question test. Three systems use different models for different input formats.
- Made by
- Charles University and Brno University of Technology
- Released
- July 2026
- Kind
- Audio understanding and voice
- Level
- University
- Scored
- Checked automatically
- Answers
- Multiple choice
- Tasks
- All public
- Measures
- Capability
- Subject
- Arts
- Language
- English
- Interaction
- Single turn
- Items from
- Another benchmark
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Audio and speech
- Strongest model tested
-
Gemini 3.1 Pro #40 on the capability index
- Last model added
- March 2026
- Scores refreshed
- November 20th, 2026
Gemini 3.1 Pro leads with 59% among models released by March 2026 Gemini 3.1 Pro leads with 59% among models released by March 2026
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 March 2026, 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
Gemini 3.1 Pro 59%
- 2
Qwen3-Omni-30B-A3B-Thinking 46.3%
- 3
Gemini 3.1 Flash-Lite 45.7%
- 4 aggregate-gpt-5 42%
- 5 aggregate-gpt-4o 41%
- 6
Gemini 2.5 Flash-Lite Sep 2025 41%
- 7 google_gemini-2.0-flash-lite-001 40.3%
- 8 aggregate-mistral 39.3%
- 9 aggregate-gpt-5-full 37%
Source: Official MusICA-MetaBench leaderboard
- 1
Gemini 3.1 Pro 59%
- 2
Qwen3-Omni-30B-A3B-Thinking 46.3%
- 3
Gemini 3.1 Flash-Lite 45.7%
- 4
Gemini 2.5 Flash-Lite Sep 2025 41%
Source: Official MusICA-MetaBench leaderboard
| # | Entry | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 59% | Feb 2026 | Official leaderboard | |
| 2 | | 46.3% | Sep 2025 | Official leaderboard | |
| 3 | | 45.7% | Mar 2026 | Official leaderboard | |
| 4 | aggregate-gpt-5 | 42% | aggregate-gpt-5 | Official leaderboard | |
| 5 | aggregate-gpt-4o | 41% | aggregate-gpt-4o | Official leaderboard | |
| 6 | | 41% | Sep 2025 | Official leaderboard | |
| 7 | google_gemini-2.0-flash-lite-001 | 40.3% | Official leaderboard | ||
| 8 | aggregate-mistral | 39.3% | aggregate-mistral | Official leaderboard | |
| 9 | aggregate-gpt-5-full | 37% | aggregate-gpt-5-full | Official leaderboard |
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 59% | Feb 2026 | Official leaderboard | |
| 2 | | 46.3% | Sep 2025 | Official leaderboard | |
| 3 | | 45.7% | Mar 2026 | Official leaderboard | |
| 4 | | 41% | Sep 2025 | Official leaderboard |
Rankings by part
We read 53 parts from the benchmark’s own leaderboard in 9 groups: submodality, piece_id, is_nota_correct, meta-question_id, category, subcategory, skill, rel/abs and specification. Each ranks models on one part of the test. Gemini 3.1 Pro leads 42 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.
audio.mastermix.wav
4 models we track have a score here.
- 1
Gemini 3.1 Pro 46%
- 2
Qwen3-Omni-30B-A3B-Thinking 33%
- 3
Gemini 3.1 Flash-Lite 33%
- 4
Gemini 2.5 Flash-Lite Sep 2025 31%
symbolic.abc.txt
4 models we track have a score here.
- 1
Gemini 3.1 Pro 86%
- 2
Qwen3-Omni-30B-A3B-Thinking 66%
- 3
Gemini 3.1 Flash-Lite 63%
- 4
Gemini 2.5 Flash-Lite Sep 2025 57%
visual.short.png
4 models we track have a score here.
- 1
Gemini 3.1 Pro 45%
- 2
Gemini 3.1 Flash-Lite ↑1 41%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 40%
- 4
Gemini 2.5 Flash-Lite Sep 2025 35%
Anonymous_AusMeinesHerzensGrunde
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 27.8%
- 2
Gemini 3.1 Pro ↓1 27.8%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 16.7%
- 4
Gemini 3.1 Flash-Lite ↓1 16.7%
Bach_IchStehAnDeinerKrippe
4 models we track have a score here.
- 1
Gemini 3.1 Pro 51.5%
- 2
Gemini 2.5 Flash-Lite Sep 2025 ↑2 39.4%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 33.3%
- 4
Gemini 3.1 Flash-Lite ↓1 30.3%
Crueger_AufAufMeinHerzMitFreuden
4 models we track have a score here.
- 1
Gemini 3.1 Pro 63.3%
- 2
Gemini 3.1 Flash-Lite ↑1 46.7%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 40%
- 4
Gemini 2.5 Flash-Lite Sep 2025 36.7%
Drese_JesuGehVoran
4 models we track have a score here.
- 1
Gemini 3.1 Pro 61.5%
- 2
Qwen3-Omni-30B-A3B-Thinking 51.3%
- 3
Gemini 3.1 Flash-Lite 51.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 38.5%
Gesius_BefiehlDuDeineWege
4 models we track have a score here.
- 1
Gemini 3.1 Pro 58.3%
- 2
Qwen3-Omni-30B-A3B-Thinking 41.7%
- 3
Gemini 3.1 Flash-Lite 41.7%
- 4
Gemini 2.5 Flash-Lite Sep 2025 20.8%
Gesius_DuFriedensfuerstHerrJesuChrist
4 models we track have a score here.
- 1
Qwen3-Omni-30B-A3B-Thinking ↑1 52.4%
- 2
Gemini 3.1 Pro ↓1 52.4%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 47.6%
- 4
Gemini 3.1 Flash-Lite ↓1 35.7%
Jan_DuGrosserSchmerzensmann
4 models we track have a score here.
- 1
Gemini 3.1 Pro 64.3%
- 2
Gemini 2.5 Flash-Lite Sep 2025 ↑2 59.5%
- 3
Gemini 3.1 Flash-Lite 54.8%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 52.4%
Telemann_DerLiebenSonneLichtUndPracht
4 models we track have a score here.
- 1
Gemini 3.1 Pro 74.1%
- 2
Gemini 3.1 Flash-Lite ↑1 74.1%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 55.6%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 51.9%
Vulpius_ChristusDerIstMeinLeben
4 models we track have a score here.
- 1
Gemini 3.1 Pro 66.7%
- 2
Qwen3-Omni-30B-A3B-Thinking 52.4%
- 3
Gemini 3.1 Flash-Lite 47.6%
- 4
Gemini 2.5 Flash-Lite Sep 2025 38.1%
Vulpius_DieHelleSonnLeuchtJetztHerfuer
4 models we track have a score here.
- 1
Gemini 3.1 Pro 62.5%
- 2
Qwen3-Omni-30B-A3B-Thinking 58.3%
- 3
Gemini 3.1 Flash-Lite 50%
- 4
Gemini 2.5 Flash-Lite Sep 2025 25%
0
4 models we track have a score here.
- 1
Gemini 3.1 Pro 58.3%
- 2
Gemini 2.5 Flash-Lite Sep 2025 ↑2 44.6%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 42.9%
- 4
Gemini 3.1 Flash-Lite ↓1 41.7%
1
4 models we track have a score here.
- 1
Gemini 3.1 Pro 61.7%
- 2
Gemini 3.1 Flash-Lite ↑1 61.7%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 60%
- 4
Gemini 2.5 Flash-Lite Sep 2025 26.7%
1
4 models we track have a score here.
- 1
Gemini 3.1 Pro 50%
- 2
Qwen3-Omni-30B-A3B-Thinking 46.7%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 33.3%
- 4
Gemini 3.1 Flash-Lite ↓1 30%
10
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 38.1%
- 2
Qwen3-Omni-30B-A3B-Thinking 33.3%
- 3
Gemini 3.1 Pro ↓2 33.3%
- 4
Gemini 3.1 Flash-Lite ↓1 28.6%
11
4 models we track have a score here.
- 1
Gemini 3.1 Pro 33.3%
- 2
Gemini 2.5 Flash-Lite Sep 2025 ↑2 27.8%
- 3
Gemini 3.1 Flash-Lite 27.8%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 22.2%
2
4 models we track have a score here.
- 1
Gemini 3.1 Pro 70%
- 2
Qwen3-Omni-30B-A3B-Thinking 60%
- 3
Gemini 3.1 Flash-Lite 56.7%
- 4
Gemini 2.5 Flash-Lite Sep 2025 53.3%
3
4 models we track have a score here.
- 1
Gemini 3.1 Pro 50%
- 2
Qwen3-Omni-30B-A3B-Thinking 36.7%
- 3
Gemini 3.1 Flash-Lite 30%
- 4
Gemini 2.5 Flash-Lite Sep 2025 20%
4
4 models we track have a score here.
- 1
Gemini 3.1 Pro 73.3%
- 2
Gemini 3.1 Flash-Lite ↑1 60%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 53.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 33.3%
5
4 models we track have a score here.
- 1
Gemini 3.1 Pro 70%
- 2
Gemini 3.1 Flash-Lite ↑1 53.3%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 43.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 43.3%
6
4 models we track have a score here.
- 1
Gemini 3.1 Pro 80%
- 2
Gemini 3.1 Flash-Lite ↑1 66.7%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 60%
- 4
Gemini 2.5 Flash-Lite Sep 2025 50%
7
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 66.7%
- 2
Gemini 3.1 Pro ↓1 66.7%
- 3
Gemini 3.1 Flash-Lite 60%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 53.3%
8
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 40%
- 2
Qwen3-Omni-30B-A3B-Thinking 26.7%
- 3
Gemini 3.1 Flash-Lite 23.3%
- 4
Gemini 3.1 Pro ↓3 20%
9
4 models we track have a score here.
- 1
Gemini 3.1 Pro 95.2%
- 2
Qwen3-Omni-30B-A3B-Thinking 66.7%
- 3
Gemini 3.1 Flash-Lite 57.1%
- 4
Gemini 2.5 Flash-Lite Sep 2025 38.1%
harmony
4 models we track have a score here.
- 1
Gemini 3.1 Pro 55%
- 2
Qwen3-Omni-30B-A3B-Thinking 41.7%
- 3
Gemini 3.1 Flash-Lite 38.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 35%
pitch
4 models we track have a score here.
- 1
Gemini 3.1 Pro 60.8%
- 2
Qwen3-Omni-30B-A3B-Thinking 49.2%
- 3
Gemini 3.1 Flash-Lite 44.2%
- 4
Gemini 2.5 Flash-Lite Sep 2025 35%
rhythm
4 models we track have a score here.
- 1
Gemini 3.1 Pro 59.2%
- 2
Gemini 3.1 Flash-Lite ↑1 50.8%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 50%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 45.8%
harmonic-recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 55%
- 2
Qwen3-Omni-30B-A3B-Thinking 41.7%
- 3
Gemini 3.1 Flash-Lite 38.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 35%
interval-recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 61.7%
- 2
Qwen3-Omni-30B-A3B-Thinking 45%
- 3
Gemini 3.1 Flash-Lite 45%
- 4
Gemini 2.5 Flash-Lite Sep 2025 26.7%
pitch-recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 60%
- 2
Qwen3-Omni-30B-A3B-Thinking 53.3%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 43.3%
- 4
Gemini 3.1 Flash-Lite ↓1 43.3%
proportion-recognition
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 53.3%
- 2
Gemini 3.1 Pro ↓1 43.3%
- 3
Gemini 3.1 Flash-Lite 41.7%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 40%
rhythmic-recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 75%
- 2
Gemini 3.1 Flash-Lite ↑1 60%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 51.7%
- 4
Gemini 2.5 Flash-Lite Sep 2025 46.7%
cadence recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 33.3%
- 2
Gemini 2.5 Flash-Lite Sep 2025 ↑2 27.8%
- 3
Gemini 3.1 Flash-Lite 27.8%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 22.2%
chord recognition
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 38.1%
- 2
Qwen3-Omni-30B-A3B-Thinking 33.3%
- 3
Gemini 3.1 Pro ↓2 33.3%
- 4
Gemini 3.1 Flash-Lite ↓1 28.6%
interval recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 50%
- 2
Qwen3-Omni-30B-A3B-Thinking 36.7%
- 3
Gemini 3.1 Flash-Lite 30%
- 4
Gemini 2.5 Flash-Lite Sep 2025 20%
interval statistics
4 models we track have a score here.
- 1
Gemini 3.1 Pro 73.3%
- 2
Gemini 3.1 Flash-Lite ↑1 60%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 53.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 33.3%
note length
4 models we track have a score here.
- 1
Gemini 3.1 Pro 70%
- 2
Gemini 3.1 Flash-Lite ↑1 53.3%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 43.3%
- 4
Gemini 2.5 Flash-Lite Sep 2025 43.3%
note length statistics
4 models we track have a score here.
- 1
Gemini 3.1 Pro 80%
- 2
Gemini 3.1 Flash-Lite ↑1 66.7%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 60%
- 4
Gemini 2.5 Flash-Lite Sep 2025 50%
pitch recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 50%
- 2
Qwen3-Omni-30B-A3B-Thinking 46.7%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 33.3%
- 4
Gemini 3.1 Flash-Lite ↓1 30%
pitch statistics
4 models we track have a score here.
- 1
Gemini 3.1 Pro 70%
- 2
Qwen3-Omni-30B-A3B-Thinking 60%
- 3
Gemini 3.1 Flash-Lite 56.7%
- 4
Gemini 2.5 Flash-Lite Sep 2025 53.3%
rhythm pattern
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 66.7%
- 2
Gemini 3.1 Pro ↓1 66.7%
- 3
Gemini 3.1 Flash-Lite 60%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 53.3%
rhythm pattern statistics
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 40%
- 2
Qwen3-Omni-30B-A3B-Thinking 26.7%
- 3
Gemini 3.1 Flash-Lite 23.3%
- 4
Gemini 3.1 Pro ↓3 20%
tonality recognition
4 models we track have a score here.
- 1
Gemini 3.1 Pro 95.2%
- 2
Qwen3-Omni-30B-A3B-Thinking 66.7%
- 3
Gemini 3.1 Flash-Lite 57.1%
- 4
Gemini 2.5 Flash-Lite Sep 2025 38.1%
absolute
4 models we track have a score here.
- 1
Gemini 3.1 Pro 62.4%
- 2
Qwen3-Omni-30B-A3B-Thinking 49.6%
- 3
Gemini 3.1 Flash-Lite 48.2%
- 4
Gemini 2.5 Flash-Lite Sep 2025 44%
relative
4 models we track have a score here.
- 1
Gemini 3.1 Pro 56%
- 2
Qwen3-Omni-30B-A3B-Thinking 43.4%
- 3
Gemini 3.1 Flash-Lite 43.4%
- 4
Gemini 2.5 Flash-Lite Sep 2025 38.4%
harmony-absolute
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 38.1%
- 2
Qwen3-Omni-30B-A3B-Thinking 33.3%
- 3
Gemini 3.1 Pro ↓2 33.3%
- 4
Gemini 3.1 Flash-Lite ↓1 28.6%
harmony-relative
4 models we track have a score here.
- 1
Gemini 3.1 Pro 66.7%
- 2
Qwen3-Omni-30B-A3B-Thinking 46.1%
- 3
Gemini 3.1 Flash-Lite 43.6%
- 4
Gemini 2.5 Flash-Lite Sep 2025 33.3%
pitch-absolute
4 models we track have a score here.
- 1
Gemini 3.1 Pro 60%
- 2
Qwen3-Omni-30B-A3B-Thinking 53.3%
- 3
Gemini 2.5 Flash-Lite Sep 2025 ↑1 43.3%
- 4
Gemini 3.1 Flash-Lite ↓1 43.3%
pitch-relative
4 models we track have a score here.
- 1
Gemini 3.1 Pro 61.7%
- 2
Qwen3-Omni-30B-A3B-Thinking 45%
- 3
Gemini 3.1 Flash-Lite 45%
- 4
Gemini 2.5 Flash-Lite Sep 2025 26.7%
rhythm-absolute
4 models we track have a score here.
- 1
Gemini 3.1 Pro 75%
- 2
Gemini 3.1 Flash-Lite ↑1 60%
- 3
Qwen3-Omni-30B-A3B-Thinking ↓1 51.7%
- 4
Gemini 2.5 Flash-Lite Sep 2025 46.7%
rhythm-relative
4 models we track have a score here.
- 1
Gemini 2.5 Flash-Lite Sep 2025 ↑3 53.3%
- 2
Gemini 3.1 Pro ↓1 43.3%
- 3
Gemini 3.1 Flash-Lite 41.7%
- 4
Qwen3-Omni-30B-A3B-Thinking ↓2 40%
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official MusICA-MetaBench 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 MusICA-MetaBench 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 Audio and video
- Video-MMMU In use 300 expert-level lecture videos with questions testing perception, comprehension and adaptation of knowledge. Accuracy.
Qwen 3.8 Max 88.7%
- LVBench In use Hours-long video understanding with multiple-choice questions on reasoning, grounding and summarization.
Gemini 4 Argon 91.7%
- MMVU In use 3,000 expert-annotated questions on videos across 27 disciplines. Multiple-choice and open-ended accuracy.
Qwen 3.8 Max 82.4%
- MLVU In use Multi-task long video understanding (minutes to hours) with multiple-choice and generation tasks. Accuracy.
Qwen 3.8 Max 90.8%
- FLEURS In use Models transcribe read speech in 102 languages built from FLoRes-101 sentences. Model cards usually report word error rate, so lower scores are better.
Phi-4-multimodal 4.00%
- VoiceBench In use Spoken instructions with speaker, environment and content variation. Voice assistant responses scored by GPT judge or rules.
Qwen3.5-Omni-Plus 93.1%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 8 months ago, in March 2026, for Gemini 3.1 Flash-Lite. 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 (9 entries).
- Pass Models scored 4 models we track, plus 5 other entries on its leaderboard.
- Warning Newest score The newest came 8 months ago, in March 2026, for Gemini 3.1 Flash-Lite.
- Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 3.1 Pro (#40).
- Pass Headroom The best is 59%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on github.com: 4 rows for models we track and 5 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.