Our new X account is live! Follow @wizwand_team for updates

Multimodal Reasoning on MathVista (Accuracy, AvgLen, Ratio)

85.9Accuracy

Qwen3-VL-32B-Thinking

Updated 3d ago

Evaluation Results

Method	Links
Qwen3-VL-32B-Thinking 2026.01		85.9	-	-
Qwen3-VL-8B-Thinking 2026.01		81.4	-	-
EvoCUA-32B 2026.01		80.4	-	-
EvoCUA-8B 2026.01		75.8	-	-
ThinkLite-VL-7B 2025.12		71.9	-	-
MM-Eureka-Qwen-7B 2025.12		71.2	-	-
MSSR 2025.12		71.1	-	-
OpenCUA-72B 2026.01		70.9	-	-
OpenVLThinker-7B 2025.12		70	-	-
GRPO 2025.12		70	-	-
EvoCUA-OpenCUA-72B 2026.01		69.4	-	-
RLOO 2025.12		69.2	-	-
REINFORCE++ 2025.12		68.5	-	-
VLAA-Thinker-7B 2025.12		68	-	-
Qwen2.5-VL-7B 2025.12		67.2	-	-
MSSR 2025.12		63	-	-
R1-Onevision-7B 2025.12		62.9	-	-
GRPO 2025.12		61.7	-	-
RLOO 2025.12		59.7	-	-
Qwen2.5-VL-3B 2025.12		59.5	-	-
REINFORCE++ 2025.12		47.7	-	-
Prune-on-Logic 2026.02		45.9	299.5	6.5
No Compression 2026.02		45.7	511.6	11.2
XMCC 2026.02		45.3	91	2
StepEntropy 2026.02		43.9	251.3	5.7
XMCC 2026.02		37.6	89.7	2.4
Prune-on-Logic 2026.02		37.3	324.2	8.7
StepEntropy 2026.02		36.8	232.2	6.3
No Compression 2026.02		35.8	524.1	14.6