Leaderboard

Compare human–model fit on the open public set, the licensed semi-public research set, or the complete CogGym benchmark. Choose R² or normalized distributional divergence, then rank models by text, image, or video performance. A dash indicates no available score.

Browse Experiments

Metric

Rank by modality

Full 258-experiment benchmark (91 text, 82 image, 85 video); available experiment counts vary by model. Human split-half R² is Spearman–Brown corrected. Higher is better. Models are ranked by text R² only when scored on at least half of the available experiments for that modality; lower-coverage results are shown without a rank.

RankModelText R²Image R²Video R²

—

Human split-halfHuman baseline

Spearman–Brown corrected; archival reference: 80 text, 33 image, 59 video experiments.

0.93

0.95

0.92

GPT-5.6 SolOpenAI

85 text, 80 image experiments scored

0.61

0.63

Gemini 3.1 ProGoogle

91 text, 82 image, 84 video experiments scored

0.58

0.42

Claude Opus 4.7Anthropic

91 text, 80 image experiments scored

0.55

0.45

GLM 5.2Z.ai

88 text experiments scored

0.53

Claude Sonnet 4.5Anthropic

91 text, 80 image experiments scored

0.52

0.38

GPT-4oOpenAI

91 text, 81 image experiments scored

0.52

0.29

Gemini 3 FlashGoogle

91 text, 80 image, 85 video experiments scored

0.52

0.28

Gemini 2.5 ProGoogle

91 text, 81 image, 84 video experiments scored

0.51

0.53

0.25

Gemini 2.5 FlashGoogle

91 text, 81 image, 84 video experiments scored

0.50

0.47

0.19

#10

Grok 4.3xAI

87 text, 80 image experiments scored

0.50

0.30

#11

Qwen3-VL 32BAlibaba

91 text, 80 image, 84 video experiments scored

0.49

0.25

0.11

#12

GPT-4.1OpenAI

91 text, 80 image experiments scored

0.49

0.32

#13

DeepSeek V4 ProDeepSeek

90 text experiments scored

0.49

#14

Llama 3.1 70BMeta

87 text experiments scored

0.48

#15

Qwen3.5 Flash 35BAlibaba

89 text, 81 image, 47 video experiments scored

0.48

0.39

0.12

#16

GPT-5 miniOpenAI

91 text, 81 image experiments scored

0.47

0.46

#17

Kimi K2.5Moonshot

90 text, 81 image experiments scored

0.47

0.33

#18

GPT-OSS 120BOpenAI

90 text experiments scored

0.47

#19

Qwen2.5-VL 32BAlibaba

88 text, 80 image, 84 video experiments scored

0.47

0.24

0.08

#20

Llama 3.3 70BMeta

88 text experiments scored

0.46

#21

GPT-OSS 20BOpenAI

90 text experiments scored

0.46

#22

DeepSeek V4 FlashDeepSeek

90 text experiments scored

0.46

#23

Claude Haiku 4.5Anthropic

91 text, 81 image experiments scored

0.46

0.22

#24

Mistral Small 24BMistral

91 text experiments scored

0.45

#25

Gemma 4 26BGoogle

91 text, 81 image experiments scored

0.45

0.27

#26

GPT-4o miniOpenAI

91 text, 81 image experiments scored

0.44

0.21

#27

Gemma 3 27BGoogle

91 text, 81 image experiments scored

0.43

0.21

#28

Llama 4 MaverickMeta

91 text, 81 image experiments scored

0.41

0.23

#29

DeepSeek V3.2DeepSeek

88 text experiments scored

0.40

#30

Llama 4 ScoutMeta

91 text, 81 image experiments scored

0.40

0.21

#31

Gemma 3 12BGoogle

90 text experiments scored

0.37

#32

Qwen3-VL 8BAlibaba

85 text, 70 image, 81 video experiments scored

0.37

0.19

0.08

#33

Qwen3.5 9BAlibaba

91 text, 80 image, 84 video experiments scored

0.36

0.19

0.10

#34

Llama 3.1 8BMeta

87 text experiments scored

0.34

#35

Qwen3-VL 4BAlibaba

88 text, 70 image, 80 video experiments scored

0.33

0.14

0.08

#36

Qwen2.5-VL 7BAlibaba

89 text, 80 image, 84 video experiments scored

0.33

0.12

0.07

#37

Llama 3.2 11BMeta

87 text, 11 image experiments scored

0.32

0.22

#38

Qwen3.5 4BAlibaba

90 text, 80 image, 84 video experiments scored

0.30

0.19

0.09

#39

Mistral 7BMistral

89 text experiments scored

0.29

#40

Gemma 4 E2BGoogle

89 text experiments scored

0.27

#41

Llama 3.2 3BMeta

90 text experiments scored

0.26

#42

Qwen2.5-VL 3BAlibaba

90 text, 80 image, 84 video experiments scored

0.25

0.10

0.05

#43

Gemma 3 4BGoogle

90 text experiments scored

0.24

#44

CentaurResearch model

91 text experiments scored

0.22

#45

Qwen3 1.7BAlibaba

84 text experiments scored

0.21

#46

Qwen2.5 1.5BAlibaba

85 text experiments scored

0.14

#47

Qwen3.5 2BAlibaba

89 text, 80 image, 84 video experiments scored

0.12

0.08

0.05

#48

Qwen2.5 0.5BAlibaba

89 text experiments scored

0.11

#49

Llama 3.2 1BMeta

90 text experiments scored

0.10

#50

Qwen3.5 0.8BAlibaba

90 text, 76 image, 82 video experiments scored

0.08

0.06

0.05

Source: CogGym overall set, subset membership dated 2026-10-01

Leaderboard

Browse Experiments

Metric

Rank by modality

RankModelText R²Image R²Video R²

—

Human split-halfHuman baseline

Spearman–Brown corrected; archival reference: 80 text, 33 image, 59 video experiments.

0.93

0.95

0.92

GPT-5.6 SolOpenAI

85 text, 80 image experiments scored

0.61

0.63

Gemini 3.1 ProGoogle

91 text, 82 image, 84 video experiments scored

0.58

0.42

Claude Opus 4.7Anthropic

91 text, 80 image experiments scored

0.55

0.45

GLM 5.2Z.ai

88 text experiments scored

0.53

Claude Sonnet 4.5Anthropic

91 text, 80 image experiments scored

0.52

0.38

GPT-4oOpenAI

91 text, 81 image experiments scored

0.52

0.29

Gemini 3 FlashGoogle

91 text, 80 image, 85 video experiments scored

0.52

0.28

Gemini 2.5 ProGoogle

91 text, 81 image, 84 video experiments scored

0.51

0.53

0.25

Gemini 2.5 FlashGoogle

91 text, 81 image, 84 video experiments scored

0.50

0.47

0.19

#10

Grok 4.3xAI

87 text, 80 image experiments scored

0.50

0.30

#11

Qwen3-VL 32BAlibaba

91 text, 80 image, 84 video experiments scored

0.49

0.25

0.11

#12

GPT-4.1OpenAI

91 text, 80 image experiments scored

0.49

0.32

#13

DeepSeek V4 ProDeepSeek

90 text experiments scored

0.49

#14

Llama 3.1 70BMeta

87 text experiments scored

0.48

#15

Qwen3.5 Flash 35BAlibaba

89 text, 81 image, 47 video experiments scored

0.48

0.39

0.12

#16

GPT-5 miniOpenAI

91 text, 81 image experiments scored

0.47

0.46

#17

Kimi K2.5Moonshot

90 text, 81 image experiments scored

0.47

0.33

#18

GPT-OSS 120BOpenAI

90 text experiments scored

0.47

#19

Qwen2.5-VL 32BAlibaba

88 text, 80 image, 84 video experiments scored

0.47

0.24

0.08

#20

Llama 3.3 70BMeta

88 text experiments scored

0.46

#21

GPT-OSS 20BOpenAI

90 text experiments scored

0.46

#22

DeepSeek V4 FlashDeepSeek

90 text experiments scored

0.46

#23

Claude Haiku 4.5Anthropic

91 text, 81 image experiments scored

0.46

0.22

#24

Mistral Small 24BMistral

91 text experiments scored

0.45

#25

Gemma 4 26BGoogle

91 text, 81 image experiments scored

0.45

0.27

#26

GPT-4o miniOpenAI

91 text, 81 image experiments scored

0.44

0.21

#27

Gemma 3 27BGoogle

91 text, 81 image experiments scored

0.43

0.21

#28

Llama 4 MaverickMeta

91 text, 81 image experiments scored

0.41

0.23

#29

DeepSeek V3.2DeepSeek

88 text experiments scored

0.40

#30

Llama 4 ScoutMeta

91 text, 81 image experiments scored

0.40

0.21

#31

Gemma 3 12BGoogle

90 text experiments scored

0.37

#32

Qwen3-VL 8BAlibaba

85 text, 70 image, 81 video experiments scored

0.37

0.19

0.08

#33

Qwen3.5 9BAlibaba

91 text, 80 image, 84 video experiments scored

0.36

0.19

0.10

#34

Llama 3.1 8BMeta

87 text experiments scored

0.34

#35

Qwen3-VL 4BAlibaba

88 text, 70 image, 80 video experiments scored

0.33

0.14

0.08

#36

Qwen2.5-VL 7BAlibaba

89 text, 80 image, 84 video experiments scored

0.33

0.12

0.07

#37

Llama 3.2 11BMeta

87 text, 11 image experiments scored

0.32

0.22

#38

Qwen3.5 4BAlibaba

90 text, 80 image, 84 video experiments scored

0.30

0.19

0.09

#39

Mistral 7BMistral

89 text experiments scored

0.29

#40

Gemma 4 E2BGoogle

89 text experiments scored

0.27

#41

Llama 3.2 3BMeta

90 text experiments scored

0.26

#42

Qwen2.5-VL 3BAlibaba

90 text, 80 image, 84 video experiments scored

0.25

0.10

0.05

#43

Gemma 3 4BGoogle

90 text experiments scored

0.24

#44

CentaurResearch model

91 text experiments scored

0.22

#45

Qwen3 1.7BAlibaba

84 text experiments scored

0.21

#46

Qwen2.5 1.5BAlibaba

85 text experiments scored

0.14

#47

Qwen3.5 2BAlibaba

89 text, 80 image, 84 video experiments scored

0.12

0.08

0.05

#48

Qwen2.5 0.5BAlibaba

89 text experiments scored

0.11

#49

Llama 3.2 1BMeta

90 text experiments scored

0.10

#50

Qwen3.5 0.8BAlibaba

90 text, 76 image, 82 video experiments scored

0.08

0.06

0.05

Source: CogGym overall set, subset membership dated 2026-10-01

Leaderboard

Loading route

Leaderboard