List your company
ResearchFormerly WizWand

Multimodal Reasoning on MMMU

83.89AccuracyGemini-2.5 (Pro)
57.588464.416771.24578.0733Feb 2, 2023Aug 22, 2023Mar 10, 2024Sep 28, 2024Apr 17, 2025Nov 4, 2025May 25, 2026
Updated 4mo ago

Evaluation results

MethodLinks
Gemini-2.5 (Pro)Model Tier=Pro2026.0183.89
S1-VL-32B-RLParameter Count=32B, Training Phase=RL2026.0483.4
S1-VL-32B-SFTParameter Count=32B, Training Phase=SFT2026.0482.5
Claude Sonnet 4.5Input Modality=Multimodal, LLM-as-a-Judge=GPT-4o2025.1182
Gemini 2.5 Pro2026.0482
Qwen3.5Parameters=35BA3B, Internal Reasoning (Think mode)=false2026.0581.4
GPT-52026.0481.22
SenseNova-U1Parameters=30BA3B, Internal Reasoning (Think mode)=true2026.0580.55
STEP3-VL-10B (PaCoRe)Reasoning Strategy=PaCoRe, Parameters=10B2026.0180.11
Seed-1.5-VL (Thinking)Thinking Mode=true2026.0179.11
Gemini 2.5 ProInput Modality=Multimodal, LLM-as-a-Judge=GPT-4o2025.1179
Qwen3-VL (Thinking)Thinking Mode=true, Parameters=235B-A22B2026.0178.7
Qwen3.5Parameters=9B, Internal Reasoning (Think mode)=false2026.0578.4
STEP3-VL-10BNumber of Parameters=10B2026.0178.11
STEP3-VL-10B (SeRe)Reasoning Strategy=SeRe, Parameters=10B2026.0178.11
Qwen3-VL-235B-A22B-ThinkingParameter Count=235B-A22B, Reasoning Mode=Thinking2026.0477.89
GPT-5 highInput Modality=Multimodal, LLM-as-a-Judge=GPT-4o2025.1177
Gemma4Parameters=26BA4B, Internal Reasoning (Think mode)=false2026.0576.56
Qwen3-VL-32B-ThinkingParameter Count=32B, Reasoning Mode=Thinking2026.0476
Qwen3VLParameters=30BA3B, Internal Reasoning (Think mode)=true2026.0576
Athena-PRMPolicy Model=Qwen2.5-VL-72B, Best-of-N Evaluation=82025.0675.8
Intern-S1Parameter Count=235B+6B2026.0475.56
GLM-4.6VParameters=106B-A12B2026.0175.2
Claude-3.7-SonnetModel Category=Proprietary2025.0675
SenseNova-U1Parameters=8B, Internal Reasoning (Think mode)=true2026.0574.78
Qwen3VLParameters=8B, Internal Reasoning (Think mode)=true2026.0574.1
Qwen3-VL ThinkingNumber of Parameters=8B2026.0173.53
Gemini 2.5 Flash2026.0472.7
Gemini-2.0-FlashModel Category=Proprietary2025.0672.6
Qwen2.5-VL-72B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=GPT-OSS-120B-A5B2025.0672.4
Athena-ORMPolicy Model=Qwen2.5-VL-72B, Best-of-N Evaluation=82025.0672.3
InternVL3-78BModel Category=Open-Source2025.0672.2
InternVL 3.5Number of Parameters=8B2026.0171.69
GLM-4.6V FlashNumber of Parameters=9B2026.0171.17
MiMo-VL RL-2508Number of Parameters=7B2026.0171.14
Self-consistencyPolicy Model=Qwen2.5-VL-72B, Best-of-N Evaluation=82025.0671.1
GPT-4o-20241120Model Category=Proprietary2025.0670.7
Gemini-2.0-FlashActivation Replay=false2025.1170.7
LongCat-NextParameters=68BA3B, Internal Reasoning (Think mode)=false2026.0570.6
QVQ-72B-PreviewModel Category=Open-Source2025.0670.3
QvQ-72B-PreviewActivation Replay=false2025.1170.3
Qwen2.5-VL-72BPolicy Model=Qwen2.5-VL-72B, Best-of-N Evaluation=82025.0670.2
InternVL3-38BModel Category=Open-Source2025.0670.1
Intern-S1-miniParameter Count=8B2026.0470
Qwen2.5-VL-32B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=GPT-OSS-120B-A5B2025.0669.8
ICLAModel=Qwen2.5-VL-7B2026.0269.2
Sora-2 AudioInput Modality=Audio, LLM-as-a-Judge=GPT-4o2025.1169.2
Qwen2.5-VL-32BReasoning Framework=Standard2025.0669
VCDModel=Qwen2.5-VL-7B2026.0268.3
Qwen2.5-VL-72BModel Category=Open-Source2025.0668.2
Qwen2.5-VL-72BReasoning Framework=Standard2025.0668.2
Qwen2.5-VL-32B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=Qwen3-8B2025.0667.8
VanillaModel=Qwen2.5-VL-7B2026.0267.5
RLSDBase Model=Qwen3-VL-8B-Instruct, Max Context Length=81922026.0467.22
VL-Rethinker-72BModel Category=Open-Source2025.0667.2
InternVL3-14BModel Category=Open-Source2025.0667.1
Ovis2-34BModel Category=Open-Source2025.0666.7
VDDModel=Qwen2.5-VL-7B2026.0265.8
DAMOModel=Qwen2.5-VL-7B2026.0265.8
GRPOBase Model=Qwen3-VL-8B-Instruct, Max Context Length=81922026.0465.11
SDPOBase Model=Qwen3-VL-8B-Instruct, Max Context Length=81922026.0465.11
Gemma-3-27BParameters=27B2025.1264.9
Gemma-3-27BModel Category=Open-Source2025.0664.9
Qwen2.5-VL-7B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=Qwen3-8B2025.0664.7
MiMo-VL-7BBackbone Model=MiMo-VL-7B, Training Paradigm (+PC)=No2026.0364.6
OPSDBase Model=Qwen3-VL-8B-Instruct, Max Context Length=81922026.0463.82
Athena-PRMPolicy Model=Qwen2.5-VL-7B, Best-of-N Evaluation=82025.0663.8
GRPO + OPSDBase Model=Qwen3-VL-8B-Instruct, Max Context Length=81922026.0463.22
MM-Eureka-Qwen-32BActivation Replay=true2025.1163.2
InternVL3-8BParameters=8B2025.1262.7
InternVL3-8BModel Category=Open-Source2025.0662.7
InternVL3-8BActivation Replay=false2025.1162.7
Athena-ORMPolicy Model=Qwen2.5-VL-7B, Best-of-N Evaluation=82025.0662.7
MiMo-VL-7B +PCBackbone Model=MiMo-VL-7B, Training Paradigm (+PC)=Yes2026.0362.6
DeCoModel=Qwen2.5-VL-7B2026.0262.5
Base LLMBase Model=Qwen3-VL-8B-Instruct, Max Context Length=81922026.0462.44
OmniCaptionerMethod Category=Caption-then-Reason2025.0662.2
MM-Eureka-32BModel Category=Open-Source2025.0662
MM-Eureka-Qwen-7BActivation Replay=true2025.1162
ECSOMethod Category=Caption-then-Reason2025.0661.4
Qwen2.5-VL-3B w/ RAPIDReasoning Framework=RAPID, Reasoner Model=Qwen3-8B2025.0660.9
DoLAModel=Qwen2.5-VL-7B2026.0260.8
Ovis2-16BModel Category=Open-Source2025.0660.7
Athena-PRMPolicy Model=InternVL2.5-8B, Best-of-N Evaluation=82025.0660.3
VisualPRM-8BPolicy Model=InternVL2.5-8B, Best-of-N Evaluation=82025.0660.2
Self-consistencyPolicy Model=Qwen2.5-VL-7B, Best-of-N Evaluation=82025.0660.1
VL-Rethinker-7BActivation Replay=true2025.1160
Gemini Ultra2023.0259.4
MM-Eureka-Qwen-32BActivation Replay=false2025.1159.3
Athena-ORMPolicy Model=InternVL2.5-8B, Best-of-N Evaluation=82025.0659.1
InternVL3.5-2B2025.1259
InternVL3.5Number of Parameters=2.3B2025.1259
InternVL3.5#Params=2B2026.0359
MMR1-Math-v0-7BActivation Replay=true2025.1158.7
MM-Eureka-Qwen-7BActivation Replay=false2025.1158.7
VL-Rethinker-7BActivation Replay=false2025.1158.7
SPARK-VL-7BEvaluation Source=original paper2026.0558.7
Qwen2.5-VL-7BActivation Replay=false2025.1158.6
Qwen2.5Type=AR2026.0558.6
Qwen2.5-VL-7B-InstructEvaluation Source=original paper2026.0558.6