AudioMC
Multiple-choice questions asked and answered in audio. Scale AI.
As of 19 Sept 2026, Gemini 3.8 Flash leads AudioMC on BenchLeader with 60.4%, ahead of Inkling Small at 54.9%, across 27 model configurations with a published result.
- Published by
- Scale AI SEAL
- Category
- Multimodal
- Index weight
- Reference only
- Models
- 27
- Data as of
- 19 Sept 2026
Scale AI SEAL Leaderboards.
What the test looks like
Questions are spoken, sometimes with audio evidence, and the model must answer from listening.
How it is scored
Accuracy, published by Scale AI.
What to keep in mind
Only audio-capable models are listed.
- 1Gemini 3.8 Flash (high)60.4%
- 2Inkling Small54.9%
- 3Gemini 3 Pro (thinking)54.6%
- 4GPT Realtime 2 (xhigh)48.5%
- 5Gemini 2.5 Pro (thinking)46.9%
- 6Tml Interaction Small43.4%
- 7Gemini 2.5 Flash (thinking)40.0%
- 8GPT Realtime 237.6%
- 9Gemini 3.1 Flash Live (thinking)36.1%
- 10GPT Realtime 1.534.7%
- 11Gemini 3.1 Flash Live26.8%
- 12Voxtral Small 24B 250726.3%
- 13Gemini 2.5 Flash26.1%
- 14GPT 4o Audio25.4%
- 15Qwen3 Omni 30B A3B Instruct24.3%
27 of 27
| # | ||||
|---|---|---|---|---|
| 1 |  Gemini 3.8 FlashhighGoogle | 60.4% | 64.5 | 2026-09-09 |
| 2 | Inkling SmallThinking Machinesopen ↗ | 54.9% | 56.3 | 2026-07-30 |
| 3 | Gemini 3 ProthinkingGoogle | 54.6% | – | 2025-12-17 |
| 4 |  GPT Realtime 2xhighOpenAI | 48.5% | – | 2026-05-07 |
| 5 | Gemini 2.5 ProthinkingGoogle | 46.9% | – | 2025-12-17 |
| 6 | Tml Interaction SmallUnknown | 43.4% | – | 2026-05-11 |
| 7 | Gemini 2.5 FlashthinkingGoogle | 40.0% | 48.4 | 2025-12-17 |
| 8 | GPT Realtime 2OpenAI | 37.6% | – | 2026-05-07 |
| 9 | Gemini 3.1 Flash LivethinkingGoogle | 36.1% | – | 2026-03-26 |
| 10 | GPT Realtime 1.5OpenAI | 34.7% | – | 2026-02-23 |
| 11 | Gemini 3.1 Flash LiveGoogle | 26.8% | – | 2026-03-26 |
| 12 |  Voxtral Small 24B 2507Mistral AI | 26.3% | – | 2026-02-25 |
| 13 | Gemini 2.5 FlashGoogle | 26.1% | 52.5 | 2026-02-23 |
| 14 | GPT 4o AudioOpenAI | 25.4% | – | 2026-02-25 |
| 15 |  Qwen3 Omni 30B A3B InstructAlibabaopen ↗ | 24.3% | 39.4 | 2026-02-25 |
| 16 | GPT RealtimeOpenAI | 23.4% | – | 2025-12-17 |
| 17 | Gemini 2.5 Flash Native Audio 12 2025thinkingGoogle | 21.5% | – | 2026-02-25 |
| 18 | Mimo Audio 7BthinkingUnknown | 19.7% | – | 2025-12-17 |
| 19 | Mimo Audio 7BUnknown | 18.6% | – | 2025-12-17 |
| 20 | GPT Realtime MiniOpenAI | 16.6% | – | 2026-02-23 |
| 21 | Gemma 3n E4BGoogleopen | 15.5% | 36.4 | 2025-12-17 |
| 22 |  Phi 4 MultimodalMicrosoftopen | 15.5% | – | 2025-12-17 |
| 23 | GPT 4o Mini AudioOpenAI | 14.8% | – | 2025-12-17 |
| 24 | Gemini 2.5 Flash Native Audio 12 2025no reasoningGoogle | 13.9% | – | 2026-03-14 |
| 25 |  Kimi Audio 7BMoonshot AI | 13.7% | – | 2025-12-17 |
| 26 | Qwen2.5-Omni 7BAlibabaopen ↗ | 11.9% | – | 2025-12-17 |
| 27 |  Lfm2 Audio 1 5BLiquid AI | 9.3% | – | 2025-12-17 |
Cite as: BenchLeader, “AudioMC leaderboard”, https://www.benchleader.com/benchmarks/scale_audiomc, data as of 19 Sept 2026.
AudioMC: questions
- What does AudioMC measure?
- Questions are spoken, sometimes with audio evidence, and the model must answer from listening. Scores are reported in percent of tasks solved; higher is better.
- Which AI model leads AudioMC?
- Gemini 3.8 Flash leads AudioMC with 60.4% as of 19 Sept 2026, ahead of Inkling Small at 54.9%.
- How many models have AudioMC results?
- 27 model configurations have a AudioMC result on BenchLeader, all taken from Scale AI SEAL.
- Who runs AudioMC and how often is it updated?
- AudioMC is published by Scale AI SEAL. BenchLeader re-reads the published results every morning and records the date each result was published.
- Does AudioMC count toward the BenchLeader Index?
- No. AudioMC is shown for reference but left out of the composite index.