Qwen2.5-Omni-7B

Qwen2.5-Omni-7B places in the top three on TextVQA.

Qwen Released 27 Mar 2025 Family Qwen2.5-Omni The lab's model card

Benchmarks with a score

Ranked first, out of five or more models

In the top three

Qwen Mar 2025

Open weights Small

No Epoch Capabilities Index score yet

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.

Benchmark	Score		Rank	Source
MMLU-Pro	47%		#143 of 151	Model card
MMLU-Redux	71%		#38 of 38	Model card

Benchmark	Score	Rank	Source
MATH	71.5%	#38 of 93	Model card
GSM8K	88.7%	#44 of 87	Model card
GPQA Diamond	30.8%	#276 of 284	Model card

Benchmark	Score	Rank	Source
MBPP	73.2%	#33 of 66	Model card
HumanEval	78.7%	#53 of 95	Model card
MultiPL-E	65.8%	#15 of 18	Model card
LiveCodeBench	24.6%	#67 of 73	Model card

Benchmark	Score	Rank	Source
TextVQA	84.4%	#3 of 17	Model card
DocVQA	95.2%	#10 of 33	Model card
ChartQA	85.3%	#22 of 37	Model card
AHELM	49.5	#6 of 10	Official leaderboard
MME	2340.0	#4 of 6	Model card
MathVista	67.9%	#46 of 64	Model card
MuirBench	59.2%	#9 of 12	Model card
MM-Vet	66.8%	#6 of 8	Model card
CRPE	76.5%	#3 of 4	Model card
Video-MME	72.4%	#33 of 42	Model card
MMAU	65.6%	#11 of 14	Model card
RealWorldQA	70.3%	#24 of 30	Model card
OCRBench v2	57.8%	#16 of 20	Model card
EgoSchema	68.6%	#10 of 12	Model card
AI2D	83.2%	#37 of 44	Model card
RefCOCO / RefCOCO+ / RefCOCOg	87.7%	#16 of 19	Model card
MMMU	59.2%	#83 of 97	Model card
MMBench-EN v1.1	81.8%	#15 of 17	Model card
MMStar	64%	#27 of 30	Model card
MVBench	70.3%	#14 of 15	Model card
MMMU-Pro	36.6%	#70 of 72	Model card
MATH-Vision (MathVision)	25%	#42 of 42	Model card
Audio MC	11.9%	#16 of 16	Official leaderboard
CoVoST2	34.7	#11 of 11	Model card
VoiceBench	74.1%	#10 of 10	Model card
ODinW-13	42.4%	#9 of 9	Model card
MM-MT-Bench	6.00	#6 of 6	Model card
MME-RealWorld	61.6%	#4 of 4	Model card
OmniBench	56.1%	#2 of 2	Model card

Benchmark	Score		Rank	Source
LiveBench	29.6%		#27 of 27	Model card

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

Every result in the longer documents the lab published with it. 1 of them isn't in the launch post.