Benchmark Almanac

DCASE 2025 Audio Question Answering

Sun_Antgroup_task5_2 73.7% tops its leaderboard Kimi-Audio-7B-Instruct 52.1% leads the models we track
Not in model cards

Models answer questions about marine mammal calls, the timing of sounds, and overlapping events. Each of the three sections counts equally toward the final score.

Nine teams submitted 27 systems. The organizers also scored four baselines, including Gemini 2.0 Flash and Qwen2-Audio.

Released
December 2024
Kind
Audio understanding and voice
Level
University
Scored
Checked automatically
Answers
Multiple choice
Tasks
Public, plus a private test set
Measures
Capability
Language
English
In and out
Audio, Text → Text
Interaction
Single turn
Refresh
One fixed set
Run by
A host
Strongest model tested
Gemini 2.0 Flash Feb 2025 #147 on the capability index
Last model added
April 2025
Scores refreshed
November 20th, 2026
31
Entries on its own leaderboard, 3 models we track

Sun_Antgroup_task5_2 tops the leaderboard with 73.7% Kimi-Audio-7B-Instruct leads with 52.1% among models released by April 2025

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. The best model we track is Kimi-Audio-7B-Instruct, at 52.1%. This view shows only models we track, each at its best.

The newest model with a score here came out in April 2025, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:

  1. 1 Sun_Antgroup_task5_2 73.7%
  2. 2 Sun_Antgroup_task5_4 73.1%
  3. 3 Shi_USTC_task5_1 72.8%
  4. 4 Sun_Antgroup_task5_1 72.1%
  5. 5 Shi_USTC_task5_2 71.5%
  6. 6 Shi_USTC_task5_4 71.5%
  7. 7 Sun_Antgroup_task5_3 69.8%
  8. 8 Shi_USTC_task5_3 69.3%
  9. 9 Chen_SRCN_task5_3 64.9%
  10. 10 Grzeszczyk_SRPOL_task5_4 60.2%
  11. 11 Grzeszczyk_SRPOL_task5_2 60.1%
  12. 12 Grzeszczyk_SRPOL_task5_3 59.3%
  13. 13 Gibier_inria_task5_1 56%
  14. 14 Wijngaard_DACS_task5_4 55.3%
  15. 15 Wijngaard_DACS_task5_1 55%
  16. 16 Wijngaard_DACS_task5_2 54.7%
  17. 17 Wijngaard_DACS_task5_3 53.6%
  18. 18 Grzeszczyk_SRPOL_task5_1 52.8%
  19. 19 Kimi-Audio-7B-Instruct 52.1%
  20. 20 Gemini 2.0 Flash Feb 2025 51.2%
  21. 21 Baseline_AudioFlamingo2 50.9%
  22. 22 Chung_IND_task5_1 50.4%
  23. 23 Guan_HEU_task5_2 49.7%
  24. 24 Chung_IND_task5_2 49.2%
  25. 25 Chung_IND_task5_3 49%
  26. 26 Guan_HEU_task5_1 46.9%
  27. 27 Yin_XJTLU_task5_1 42.1%
  28. 28 Yin_XJTLU_task5_2 41.7%
  29. 29 Yin_XJTLU_task5_4 41.5%
  30. 30 Yin_XJTLU_task5_3 40.9%
  31. 31 Qwen2-Audio-7B-Instruct 37.2%

Source: Official DCASE 2025 Audio Question Answering leaderboard

Source: Official DCASE 2025 Audio Question Answering leaderboard

#EntryScoreSettingSubmitted byDateLinkSource
1 Sun_Antgroup_task5_2 12B 73.7% Sun dcase.community Official leaderboard
2 Sun_Antgroup_task5_4 20B 73.1% Sun dcase.community Official leaderboard
3 Shi_USTC_task5_1 8.1B 72.8% Song Yan dcase.community Official leaderboard
4 Sun_Antgroup_task5_1 12B 72.1% Sun dcase.community Official leaderboard
5 Shi_USTC_task5_2 7.1B 71.5% Song Yan dcase.community Official leaderboard
6 Shi_USTC_task5_4 7.1B 71.5% Song Yan dcase.community Official leaderboard
7 Sun_Antgroup_task5_3 8.2B 69.8% Sun dcase.community Official leaderboard
8 Shi_USTC_task5_3 8.1B 69.3% Song Yan dcase.community Official leaderboard
9 Chen_SRCN_task5_3 8.3B 64.9% Minjun Chen dcase.community Official leaderboard
10 Grzeszczyk_SRPOL_task5_4 8.4B 60.2% Grzeszczyk dcase.community Official leaderboard
11 Grzeszczyk_SRPOL_task5_2 8.4B 60.1% Grzeszczyk dcase.community Official leaderboard
12 Grzeszczyk_SRPOL_task5_3 8.4B 59.3% Grzeszczyk dcase.community Official leaderboard
13 Gibier_inria_task5_1 7.7B 56% Marcel Gibier dcase.community Official leaderboard
14 Wijngaard_DACS_task5_4 92.4B 55.3% Wijngaard dcase.community Official leaderboard
15 Wijngaard_DACS_task5_1 8.4B 55% Wijngaard dcase.community Official leaderboard
16 Wijngaard_DACS_task5_2 8.4B 54.7% Wijngaard dcase.community Official leaderboard
17 Wijngaard_DACS_task5_3 8.4B 53.6% Wijngaard dcase.community Official leaderboard
18 Grzeszczyk_SRPOL_task5_1 1.6B 52.8% Grzeszczyk dcase.community Official leaderboard
19 Kimi-Audio-7B-Instruct 52.1% Apr 2025 Official leaderboard
20 Gemini 2.0 Flash Feb 2025 51.2% Feb 2025 Official leaderboard
21 Baseline_AudioFlamingo2 0.3B 50.9% Official leaderboard
22 Chung_IND_task5_1 4.8B 50.4% Chung dcase.community Official leaderboard
23 Guan_HEU_task5_2 8.4B 49.7% Guan dcase.community Official leaderboard
24 Chung_IND_task5_2 4.8B 49.2% Chung dcase.community Official leaderboard
25 Chung_IND_task5_3 4.8B 49% Chung dcase.community Official leaderboard
26 Guan_HEU_task5_1 8.4B 46.9% Guan dcase.community Official leaderboard
27 Yin_XJTLU_task5_1 0.2B 42.1% Yin dcase.community Official leaderboard
28 Yin_XJTLU_task5_2 0.2B 41.7% Yin dcase.community Official leaderboard
29 Yin_XJTLU_task5_4 0.2B 41.5% Yin dcase.community Official leaderboard
30 Yin_XJTLU_task5_3 0.2B 40.9% Yin dcase.community Official leaderboard
31 Qwen2-Audio-7B-Instruct 37.2% Aug 2024 Official leaderboard
#ModelScoreSettingSubmitted byDateLinkSource
1 Kimi-Audio-7B-Instruct 52.1% Apr 2025 Official leaderboard
2 Gemini 2.0 Flash Feb 2025 51.2% Feb 2025 Official leaderboard
3 Qwen2-Audio-7B-Instruct 37.2% Aug 2024 Official leaderboard

Rankings by part

We read 3 parts from the benchmark’s own leaderboard: Part 1 Accuracy, Part 2 Accuracy and Part 3 Accuracy. Each ranks models on one part of the test. Kimi-Audio-7B-Instruct leads 2 of them, and Gemini 2.0 Flash leads the other one. Arrows show how far a model moved from its overall rank among the models we track.

Part 1 Accuracy

3 models we track have a score here.

  1. 1 Kimi-Audio-7B-Instruct 37.4%
  2. 2 Gemini 2.0 Flash Feb 2025 36.4%
  3. 3 Qwen2-Audio-7B-Instruct 27.7%

3 models have held number one on the official leaderboard since its release in December 2024

Lab claims count from the day the card came out, independent runs from the model’s release.

Qwen Google DeepMind Moonshot AI
Own leaderboard Not tested since Apr 2025

Source: Official DCASE 2025 Audio Question Answering leaderboard

Scores over time

All the scores we have, one dot each, so a model tested at several settings appears more than once.

Plot with other benchmarks
On the benchmark’s own leaderboard Best so far
0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Kimi-Audio-7B-Instruct Gemini 2.0 Flash Qwen2-Audio-7B-Instruct 0% 20% 40% 60% 80% 100% 2025 Near the ceiling (85%) Kimi-Audio-7B-Instruct Qwen2-Audio-7B-Instruct

Source: Official DCASE 2025 Audio Question Answering 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

All 695 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 19 months ago, in April 2025, for Kimi-Audio-7B-Instruct. Every benchmark

Catalog record 2 pass
  • Pass Task access Public tasks, plus a private test set.
  • Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
  • Pass Independent runs Scores from the official leaderboard (31 entries).
  • Pass Models scored 3 models we track, plus 28 other entries on its leaderboard.
  • Warning Newest score The newest came 19 months ago, in April 2025, for Kimi-Audio-7B-Instruct.
  • Note Frontier models None of the top 10 has a score. The strongest tested is Gemini 2.0 Flash (#147).
  • Pass Headroom The best is 52.1%, under 85%, so it still separates models.
Leaderboard import 3 pass
  • Pass Importer We read the board on dcase.community: 3 rows for models we track and 28 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.

How we check