List your company
ResearchFormerly WizWand

Multimodal Mathematical Reasoning on MathVista MINI

0.878AccuracyQwen3.5-27B
0.569120.649310.72950.80969Jun 8, 2025Aug 7, 2025Oct 7, 2025Dec 7, 2025Feb 6, 2026Apr 8, 2026Jun 8, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Qwen3.5-27BMode=REASONING, Architecture=Dense, # Total Params=27B, # Activated Params=27B2026.040.878
Qwen3-VL-32BMode=Thinking, Architecture=Dense, # Total Params=33B, # Activated Params=33B2026.040.859
Qwen3-VL-235B-A22BMode=Thinking, Architecture=MoE, # Total Params=236B, # Activated Params=23B2026.040.858
EXAONE 4.5 33BMode=REASONING, Architecture=Dense, # Total Params=33B, # Activated Params=33B2026.040.85
Qwen3-VL-8B-Instruct + CAREBackbone=Qwen3-VL-8B-Instruct, Method=CARE2025.120.821
MiMo-VL-7B-SFTModel Type=Instruct2025.120.818
MiMo-VL-7BBackbone Model=MiMo-VL-7B, Training Paradigm (+PC)=No2026.030.818
MiMo-VL-7B-RLModel Type=Reasoning2025.120.815
Qwen3-VL-8B-ThinkingModel Type=Reasoning2025.120.814
MiMo-VL-7B +PCBackbone Model=MiMo-VL-7B, Training Paradigm (+PC)=Yes2026.030.808
Keye-VL-1.5-8BModel Type=Reasoning2025.120.807
Qwen3-VL-4B-ThinkingModel Type=Reasoning2025.120.795
GPT-5 miniMode=REASONING: HIGH, Architecture=-, # Total Params=-, # Activated Params=-2026.040.791
InternVL3.5-8BModel Scale=8B, Optimization Method=N/A2026.060.784
MetisModel Category=Agentic Multimodal Models2026.040.78
Qwen3-VL-8B-InstructModel Type=Instruct2025.120.772
InternVL-3.5Model Size=4B2026.050.771
Qwen3-VL-8B-InstructModel Category=Open-source Models2026.040.763
Qwen2.5-VL-72B-IT#Data=/2025.060.758
Qwen3-VL-SegModel Size=4B, Training Stage=S-22026.050.755
ThinkLite-VL-7BModel Category=Text-only Reasoning Models2026.040.751
M2-ReasoningModel Type=Reasoning2025.120.75
VL-Rethinker-7BModel Category=Text-only Reasoning Models2026.040.749
Qwen2.5-VL-7B + CAREBackbone=Qwen2.5-VL-7B, Method=CARE2025.120.747
InternVL3.5-8B-InstructModel Type=Instruct2025.120.742
InternVL3.5-8B-MPOModel Type=Reasoning2025.120.742
Perception-R1-7B#Data=1.4K2025.060.742
Qwen2.5-VL-7B + GSPOBackbone=Qwen2.5-VL-7B, RL Method=GSPO2025.120.741
OpenAI-o1#Data=/2025.060.739
VideoAuto-R1Backbone=Qwen2.5-VL-7B2026.010.737
Qwen3-VL-4B-InstructModel Type=Instruct2025.120.737
Qwen3-VLModel Size=4B, Training Stage=instruct2026.050.737
Vision-R1-7BModel Type=Reasoning2025.120.735
InternVL3.5-4B (+OPD)Model Scale=4B, Optimization Method=OPD2026.060.733
GPT-5-NanoModel Type=Proprietary2025.120.731
Vision-R1-7B#Data=200K2025.060.731
InternVL3.5-4B (+GNDPO)Model Scale=4B, Optimization Method=GNDPO2026.060.73
Qwen2.5-VL-7B + DAPOBackbone=Qwen2.5-VL-7B, RL Method=DAPO2025.120.726
MM-Eureka-7BModel Category=Text-only Reasoning Models2026.040.726
MM-Eureka-7B#Data=15K2025.060.725
D2IparReasoning Strategy=D2I, Variant=par, Backbone=Qwen2.5-VL-7B2025.070.722
DeepEyesV2Model Category=Agentic Multimodal Models2026.040.719
VLAA-Thinker-7BModel Category=Text-only Reasoning Models2026.040.717
InternVL3-8BModel Category=Open-source Models2026.040.716
InternVL3.5-4B (Base (Instruct))Model Scale=4B, Optimization Method=Base (Instruct)2026.060.714
InternVL3.5-4B (+GSPO)Model Scale=4B, Optimization Method=GSPO2026.060.714
Gemini-2.0-ProModel Type=Proprietary2025.120.713
OpenVLThinker-7B#Data=25K2025.060.713
Qwen3-VLModel Size=4B, Training Stage=S-12026.050.709
VLAA-Thinker-7B#Data=25K2025.060.707
SophiaVL-R1-7B#Data=130K2025.060.706
OpenVLThinker-7B2025.070.702
DeepEyesModel Type=Reasoning2025.120.701
DeepEyesModel Category=Agentic Multimodal Models2026.040.701
ThymeModel Category=Agentic Multimodal Models2026.040.7
D2IlocReasoning Strategy=D2I, Variant=loc, Backbone=Qwen2.5-VL-7B2025.070.699
D2IjusReasoning Strategy=D2I, Variant=jus, Backbone=Qwen2.5-VL-7B2025.070.697
LLaVA-OneVision-1.5 8BModel Type=Instruct2025.120.696
Qwen2.5-VL-7B +PCBackbone Model=Qwen2.5-VL-7B, Training Paradigm (+PC)=Yes2026.030.696
Qwen2.5-VL-7B w/ GRPO†Reasoning Mode=Intuitive, Backbone=Qwen2.5-VL-7B2025.070.695
Qwen2.5-VL-7BBackbone=Qwen2.5-VL-7B2026.010.694
Qwen2.5-VL-7B + GRPOBackbone=Qwen2.5-VL-7B, RL Method=GRPO2025.120.689
Qwen2.5-VL-7BModel Type=Instruct2025.120.686
Qwen-2.5-VL-7B-InstructModel Category=Open-source Models2026.040.683
Qwen2.5-VL-7BBackbone Model=Qwen2.5-VL-7B, Training Paradigm (+PC)=No2026.030.682
Qwen2.5-VL-7B*Backbone=Qwen2.5-VL-7B2025.070.682
Qwen2.5-VL-7B-IT#Data=/2025.060.681
Qwen2.5-VL-7B w/ GRPOReasoning Mode=Deliberate, Backbone=Qwen2.5-VL-7B2025.070.681
D2DlocReasoning Strategy=D2D, Variant=loc, Backbone=Qwen2.5-VL-7B2025.070.68
FineViT-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.030.677
Claude-Sonnet-3.7Model Type=Proprietary2025.120.668
Claude-3.7-Sonnet#Data=/2025.060.668
Qwen2.5-VL-3B + CAREBackbone=Qwen2.5-VL-3B, Method=CARE2025.120.665
InternVL3.5-2B (+GNDPO)Model Scale=2B, Optimization Method=GNDPO2026.060.66
D2DjusReasoning Strategy=D2D, Variant=jus, Backbone=Qwen2.5-VL-7B2025.070.657
R1-OneVision-7B#Data=155K2025.060.65
InternVL3.5-2B (+OPD)Model Scale=2B, Optimization Method=OPD2026.060.647
InternVL2.5-8B#Data=/2025.060.644
InternVL2.5-8B2025.070.644
R1-Onevision-7B2025.070.641
GPT-4oModel Type=Proprietary2025.120.638
GPT-4o#Data=/2025.060.638
GPT-4o2025.070.638
Qwen2.5-VL-3B +PCBackbone Model=Qwen2.5-VL-3B, Training Paradigm (+PC)=Yes2026.030.633
R1-VL-7B#Data=260K2025.060.627
D2DparReasoning Strategy=D2D, Variant=par, Backbone=Qwen2.5-VL-7B2025.070.626
Qwen2.5-VL-3BBackbone Model=Qwen2.5-VL-3B, Training Paradigm (+PC)=No2026.030.623
InternVL3.5-2B (+GSPO)Model Scale=2B, Optimization Method=GSPO2026.060.623
Qwen2.5-VL-3B-InstructParams=3.75B2025.120.62
Qwen2.5-VL-3BModel Type=Instruct2025.120.62
Intern3.5-VLLanguage Backbone=Qwen3 1.7B LLM, Evaluation Framework=VLMEvalKit [69]2026.030.615
InternVL3.5-2B (Base (Instruct))Model Scale=2B, Optimization Method=Base (Instruct)2026.060.608
MobileNet-QwenParams=1.84B2025.120.606
URSA-7B#Data=3.06M2025.060.598
Aquila-VLLanguage Backbone=Qwen2.5 1.5B LLM, Evaluation Framework=VLMEvalKit [69]2026.030.593
Qwen2-VL-7B-IT#Data=/2025.060.586
LLaVA-OneVisionModel Category=Open-source Models2026.040.586
InternVL2-8B2025.070.583
Qwen2-VL-7B2025.070.582
GPT-4V2025.070.581