Benchmark Almanac

Qwen3-Omni-30B-A3B

Qwen3-Omni-30B-A3B places in the top three on FLEURS.

Qwen Released September 22nd, 2025 Family Qwen3-Omni The lab's model card Compare with others
43
Benchmarks with a score
0
Ranked first, out of five or more models
1
In the top three
Qwen Sep 2025

Qwen3-Omni-30B-A3B

Open weights Mid-size

130.3 #272 of 404
Self-Reported Index, from the lab's own numbers. No Epoch Capabilities Index score yet.
  • MMMU-Pro 57% #71 /90
  • HallusionBench 59.7% #24 /31
  • OCRBench v2 61.3% #8 /14

Where it ranks

Its best score on each benchmark, and where that ranks among every model with a score there. We rank each model once, at its best setting.

Knowledge and language

BenchmarkScoreRankSource
WritingBench 82.6 #6 of 14 Lab model card
Creative Writing v3 80.6 #6 of 12 Lab model card
Multi-IF 64% #20 of 28 Lab model card
MMLU-Redux 86.6% #33 of 45 Lab model card
Arena-Hard v2.0 55.1% #18 of 24 Lab model card
IFEval 81% #103 of 135 Lab model card
HallusionBench 59.7% #24 of 31 Lab model card

Math and science

BenchmarkScoreRankSource
GPQA Diamond 69.6% #185 of 302 Lab model card
ZebraLogic 76% #19 of 23 Lab model card
AIME 2025 65% #119 of 143 Lab model card
PolyMath 37.9% #21 of 24 Lab model card

Coding

BenchmarkScoreRankSource
MultiPL-E 81.4% #4 of 18 Lab model card

Agents and tools

BenchmarkScoreRankSource
Berkeley Function Calling Leaderboard (BFCL) 64.4% #30 of 41 Lab model card

Vision, audio and robotics

BenchmarkScoreRankSource
MMAU 77.5% #5 of 15 Lab model card
Daily-Omni 73.6% #6 of 14 Official leaderboard
MathVista 75.9% #34 of 70 Lab model card
ChartQA 86.8% #18 of 37 Lab model card
FLEURS 5.33% #3 of 6 Lab model card
VoxMem 26% #5 of 9 Official leaderboard
OCRBench v2 61.3% #8 of 14 Official leaderboard
UNO-Bench 39.6% #4 of 7 Official leaderboard
VideoMME v2 (with subtitles) 19.5% #19 of 31 Official leaderboard
CountBench 90% #18 of 28 Lab model card
MMMU 69.1% #68 of 103 Lab model card
Audio MC 24.3% #12 of 18 Official leaderboard
VoiceBench 85.5% #8 of 12 Lab model card
MM-MT-Bench 7.40 #4 of 6 Lab model card
VideoDR (with search) 37% #4 of 6 Official leaderboard
AI2D 85.2% #36 of 52 Lab model card
MusTBench 30.2% #3 of 4 Official leaderboard
MMMU-Pro 57% #71 of 90 Lab model card
MMOU 35.6% #4 of 5 Official leaderboard
MLVU 75.2% #18 of 22 Lab model card
MATH-Vision (MathVision) 56.3% #42 of 51 Lab model card
MMStar 68.5% #30 of 36 Lab model card
MMSU 69% #5 of 6 Lab model card
OmniCap-IF 71.9% #5 of 6 Official leaderboard
Video-MME 70.5% #45 of 51 Lab model card
LVBench 50.2% #28 of 31 Lab model card
Video-Holmes 57.3% #6 of 6 Lab model card
WorldSense (omnimodal) 54% #6 of 6 Lab model card
MuChoMusic 52% #4 of 4 Lab model card

Other

BenchmarkScoreRankSource
LiveBench 71.8% #10 of 27 Lab model card

Its lab's numbers put it at 130.3, number 272 of 404

The Self-Reported Index fits one score to the 24 capability results Qwen reported for this model, on the scale of the Epoch Capabilities Index. Labs choose which benchmarks to show and how to run them, so read it as the lab's claim, not an independent measurement. The 90% interval runs from 128.7 to 131.8. How we compute it.

Its lab's numbers next to 2 leaderboards

Where the benchmark's own leaderboard, or Epoch AI's run, also has this model. Gaps mostly come from different harnesses, effort levels and task subsets. Every pair.

Benchmark Lab says Board says Gap
VoiceBench 85.5% VoiceBench Overall 88.8% Leaderboard, thinking −3.3 pts
Daily-Omni 75.8% DailyOmni (Thinking variant) 73.6% Leaderboard, Omni-Modal Language Models (With Visual and Audio) +2.2 pts

Its launch post

The benchmarks the lab chose to headline, with the scores it gave.

Qwen3-Omni-30B-A3B-Instruct model card (Hugging Face) Sep 22, 2025

Its system card

Every result in the longer documents the lab published with it. 3 of them aren't in the launch post.

Qwen3-Omni Technical Report Sep 22, 2025

Capabilities