Qwen3-Omni-30B-A3B-Instruct

MERA Created at 20.11.2025 16:48

Ratings for leaderboard tasks

The table will scroll to the left

Board Result Attempted Score Coverage Place in the rating
Multi 0.454 0.511 0.889 2
Images 0.499 0.499 1 2
Audio 0.531 0.531 1 2
Video 0.357 0.535 0.667 -

Tasks

The table will scroll to the left

Task Modality Result Metric
0.719
EM JudgeScore
0.335
EM F1
0.678
EM JudgeScore
0.433
EM JudgeScore
0.493
EM JudgeScore
0.712
EM JudgeScore
0.497
EM JudgeScore
0.154
EM JudgeScore
0.651
EM JudgeScore
0.697
EM JudgeScore
0.354
EM JudgeScore
0.419
EM JudgeScore
0.401
EM JudgeScore
0.59
EM JudgeScore
0.222
EM JudgeScore
culture 0.095 / 0.29
business 0.129 / 0.417
medicine 0.094 / 0.31
social_sciences 0.119 / 0.385
fundamental_sciences 0.107 / 0.316
applied_sciences 0.137 / 0.397
0.612
EM JudgeScore
biology 0.709 / 0.673
chemistry 0.703 / 0.617
physics 0.749 / 0.703
economics 0.605 / 0.479
ru 0.489 / 0.449
all 0.586 / 0.531
0.72
EM JudgeScore
biology 0.684 / 0.509
chemistry 0.731 / 0.612
physics 0.778 / 0.717
science 0.902 / 0.78

Information about the submission

Mera version
1.0.0
Torch Version
2.8.0
The version of the codebase
7e640aa
CUDA version
12.6
Precision of the model weights
bfloat16
Seed
1234
Batch
1
Transformers version
4.57.1
The number of GPUs and their type
2 x A100
Architecture
openai-chat-completions

Team:

MERA

Name of the ML model:

Qwen3-Omni-30B-A3B-Instruct

Model size

35.0B

Model type:

Opened

SFT

Inference parameters

The size of the context:
8000