MathVision

AI benchmark results and model performance

Type: percentageMath

Recorded Results

Average Score

55.87%

Score Range

25% - 94.60%

Leading Model (lowest score)

25% - Qwen 2.5 Coder 3B

Scores Over Time

Individual benchmark scores plotted by date.

Models Using This Benchmark

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Qwen	Qwen 2.5 Coder 3B	-	25%	inferred family alias from qwen2.5-omni-7b (score=0.3000; benches=45)	Yes	Source
Qwen	Qwen 2.5 Coder 7B	-	25%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4700; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math 7B	-	25%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4767; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math 7B PRM800K	-	25%	inferred family alias from qwen2.5-omni-7b (score=0.3696; benches=45)	Yes	Source
Qwen	Qwen 2.5 Math PRM 7B	-	25%	inferred family alias from qwen2.5-omni-7b (score=0.4092; benches=45)	Yes	Source
Qwen	Qwen 2.5 Omni 3B	-	25%	inferred high-confidence family alias from qwen2.5-omni-7b (score=0.4933; benches=45)	Yes	Source
Qwen	Qwen 2.5 Omni 7B	-	25%	-	Yes	Source
Qwen	Qwen 2.5 0.5B	-	25.07%	inferred family alias from qwen2.5-vl-7b (score=0.3131; benches=32)	Yes	Source
Qwen	Qwen 2.5 1.5B	-	25.07%	inferred family alias from qwen2.5-vl-7b (score=0.3131; benches=32)	Yes	Source
Qwen	Qwen 2.5 3B	-	25.07%	inferred family alias from qwen2.5-vl-7b (score=0.3531; benches=32)	Yes	Source
Qwen	Qwen 2.5 VL 7B Instruct	-	25.07%	-	Yes	Source
Qwen	QvQ 72B Preview	-	35.90%	-	Yes	Source
Qwen	Qwen 2.5 Math 72B	-	38.10%	inferred high-confidence family alias from qwen2.5-vl-72b (score=0.4700; benches=30)	Yes	Source
Qwen	Qwen 2.5 Math PRM 72B	-	38.10%	inferred family alias from qwen2.5-vl-72b (score=0.4050; benches=30)	Yes	Source
Qwen	Qwen 2.5 Math RM 72B	-	38.10%	inferred family alias from qwen2.5-vl-72b (score=0.4092; benches=30)	Yes	Source
Qwen	Qwen 2.5 VL 72B Instruct	26 Jan 2025	38.10%	-	Yes	Source
Qwen	Qwen 2.5 Coder 32B Instruct	-	38.40%	inferred high-confidence family alias from qwen2.5-vl-32b (score=0.4641; benches=28)	Yes	Source
Qwen	Qwen 2.5 VL 32B Instruct	24 Mar 2025	38.40%	-	Yes	Source
Qwen	Qwen 2.5 VL 3B Instruct	-	38.40%	inferred high-confidence family alias from qwen2.5-vl-32b (score=0.4914; benches=28)	Yes	Source
Qwen	Qwen 3 VL 4B Instruct	-	51.60%	-	Yes	-
Qwen	Qwen 3 VL 8B Instruct	-	53.90%	-	Yes	-
Qwen	Qwen 3 4B Instruct 2507	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 4B SafeRL	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 4B Thinking 2507	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3462; benches=48)	Yes	-
Qwen	Qwen 3 Embedding 4B	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 Guard Gen 4B	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3400; benches=48)	Yes	-
Qwen	Qwen 3 Guard Stream 4B	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3371; benches=48)	Yes	-
Qwen	Qwen 3 Reranker 4B	-	60%	inferred family alias from qwen3-vl-4b-thinking (score=0.3850; benches=48)	Yes	-
Qwen	Qwen 3 VL 4B Thinking	-	60%	-	Yes	-
Qwen	Qwen 3 VL 30B A3B Instruct	-	60.20%	-	Yes	-
Qwen	Qwen 3 8B	-	62.70%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.4600; benches=50)	Yes	-
Qwen	Qwen 3 Guard Stream 8B	-	62.70%	inferred family alias from qwen3-vl-8b-thinking (score=0.3371; benches=50)	Yes	-
Qwen	Qwen 3 VL 8B Thinking	-	62.70%	-	Yes	-
Qwen	Qwen 3 VL Embedding 8B	-	62.70%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5232; benches=50)	Yes	-
Qwen	Qwen 3 VL Reranker 8B	-	62.70%	inferred high-confidence family alias from qwen3-vl-8b-thinking (score=0.5275; benches=50)	Yes	-
Qwen	Qwen 3 VL 32B Instruct	-	63.40%	-	Yes	-
Qwen	Qwen 3 VL 30B A3B Thinking	-	65.70%	-	Yes	-
Qwen	Qwen 3 VL 235B A22B Instruct	-	66.50%	-	Yes	-
Qwen	Qwen 3 VL 32B Thinking	-	70.20%	-	Yes	-
Google	DiffusionGemma 26B A4B IT	10 Jun 2026	70.50%	-	Yes	Source
Qwen	Qwen 3 VL 235B A22B Thinking	-	74.60%	-	Yes	-
ByteDance	Seed 2.0 Mini	14 Feb 2026	78.10%	Seed2 official benchmark table \| MathVision	Yes	Source
Qwen	Qwen 3.5 35B A3B	24 Feb 2026	83.90%	Highest reported variant uses fixed boxed-answer prompt.	Yes	Source
Moonshot	Kimi K2.5	27 Jan 2026	84.20%	-	Yes	Source
Qwen	Qwen 3.5 27B	24 Feb 2026	86%	Highest reported variant uses fixed boxed-answer prompt.	Yes	Source
Qwen	Qwen 3.5 Flash	23 Feb 2026	86%	inferred family alias from qwen3.5-27b (score=0.4147; benches=81)	Yes	Source
Qwen	Qwen 3.5 122B A10B	24 Feb 2026	86.20%	Highest reported variant uses fixed boxed-answer prompt.	Yes	Source
ByteDance	Seed 2.0 Lite	14 Feb 2026	86.40%	Seed2 official benchmark table \| MathVision	Yes	Source
Qwen	Qwen 3.6 Plus	01 Apr 2026	88%	-	Yes	Source
ByteDance	Seed 2.0 Pro	14 Feb 2026	88.80%	Seed2 official benchmark table \| MathVision	Yes	Source
Moonshot	Kimi K3	16 Jul 2026	94.30%	Average of three runs.	Yes	Source
Qwen	Qwen 3.8 27B	14 Aug 2026	94.60%	With CI; without CI: 90.0	Yes	Source