Read charts, interpret images, and answer questions about what a picture shows.
- #1Claude Opus 5.541.57
- #2Claude Sonnet 5.541.40
- #3Claude Fable 5.141.34
- #4GPT-6 Astra40.58
- #5Claude Opus 537.84
View full vision ranking →Solve problems that need several steps, logical thinking, and careful decisions.
- #1Claude Opus 5.559.33
- #2GPT-6 Astra57.13
- #3Claude Sonnet 5.557.07
- #4Claude Opus 553.96
- #5Gemini 4 Argon53.03
View full reasoning ranking →Compare all-round ability across a broad range of tasks.
- #1Claude Opus 5.560.33
- #2Claude Sonnet 5.558.30
- #3GPT-6 Astra58.07
- #4Gemini 4 Argon54.98
- #5Claude Opus 553.74
View full general intelligence ranking →Use tools, plan steps, and carry out tasks toward a goal.
- #1Claude Opus 5.545.42
- #2GPT-6 Astra44.93
- #3Claude Sonnet 5.543.21
- #4Gemini 4 Argon41.06
- #5GPT-5.6 Sol39.12
View full agents ranking →Work with text and other inputs together, such as images, audio, or video.
- #1Claude Opus 5.541.16
- #2Claude Sonnet 5.541.13
- #3Qwen 3.8 Max38.78
- #4GPT-6 Astra37.85
- #5Kimi K337.21
View full multimodal ranking →Choose the right tool and provide the arguments it needs.
- #1GPT-6 Astra36.28
- #2Claude Sonnet 5.535.24
- #3Claude Opus 5.533.84
- #4Gemini 4 Argon31.81
- #5DeepSeek V4.1 Flash31.75
View full tool calling ranking →Write code, fix bugs, and make changes across a software project.
- #1Claude Opus 5.549.72
- #2Claude Sonnet 5.547.46
- #3GPT-6 Astra46.23
- #4Claude Fable 542.66
- #5GPT-5.6 Sol42.38
View full coding ranking →Solve equations, work with numbers, and reason through mathematical problems.
- #1Claude Fable 5.144.03
- #2Claude Opus 543.01
- #3Qwen 3.7 Max42.86
- #4Gemini 3.1 Pro Preview42.32
- #5Claude Opus 5.541.85
View full math ranking →Answer questions about living systems, from cells to ecosystems.
- #1Gemini 3.1 Pro Preview41.59
- #2GPT-6 Astra41.59
- #3GPT-5.6 Sol40.56
- #4Claude Opus 4.739.04
- #5GPT-5.538.48
View full biology ranking →Reason about molecules, reactions, and chemical properties.
- #1Gemini 3.1 Pro Preview41.87
- #2GPT-6 Astra41.59
- #3GPT-5.6 Sol40.56
- #4Qwen 3.7 Max40.43
- #5Claude Opus 4.739.04
View full chemistry ranking →Apply physical principles to explain phenomena and solve problems.
- #1Gemini 3.1 Pro Preview41.87
- #2GPT-6 Astra41.59
- #3Qwen 3.7 Max41.08
- #4GPT-5.6 Sol40.56
- #5Claude Opus 4.739.04
View full physics ranking →Find relevant information to answer a question or complete a task.
- #1Kimi K334.66
- #2GPT-6 Astra32.91
- #3Claude Opus 530.47
- #4GPT-5.6 Sol29.48
- #5GPT-5.5 Pro28.60
View full search ranking →Reason about shapes, positions, distances, and how objects move.
- #1GPT-6 Astra36.34
- #2Qwen 3.8 Max32.89
- #3Qwen 3.8 Flash31.63
- #4Qwen 3.7 Plus30.53
- #5GPT-5.6 Sol30.22
View full spatial reasoning ranking →Find and connect information across long documents or conversations.
- #1GPT-5.6 Sol32.05
- #2Gemini 4 Argon30.56
- #3Gemini 3.7 Flash30.47
- #4Qwen 3.8 Max29.06
- #5Seed 2.1 Turbo28.94
View full long context ranking →Reason through medical knowledge and health-related questions.
- #1Qwen 3.7 Max44.82
- #2Qwen 3.7 Plus42.10
- #3Qwen 3.6 Plus Uncensored42.00
- #4Qwen 3.5 397B39.17
- #5Qwen 3.6 27B36.11
View full healthcare ranking →Build web interfaces and work with layouts, styles, and interactions.
- #1Claude Fable 535.81
- #2Claude Opus 4.832.61
- #3Claude Opus 4.731.81
- #4Claude Sonnet 531.07
- #5Claude Opus 4.530.38
View full frontend development ranking →Handle risky requests and avoid harmful responses in safety evaluations.
- #1GPT-6 Astra33.19
- #2GPT-5.6 Sol29.81
- #3DeepSeek V4.1 Flash24.47
- #4GPT-5.6 Terra23.69
- #5GLM 5.322.50
View full safety ranking →Understand meaning, grammar, and how language is used.
- #1Qwen 3.7 Max41.57
- #2Qwen 3.7 Plus39.37
- #3Qwen 3.6 Plus Uncensored38.51
- #4Qwen 3.5 397B36.80
- #5NVIDIA Nemotron 3 Ultra34.36
View full language ranking →Apply scientific knowledge and reasoning across different disciplines.
- #1GPT-6 Astra20.61
- #2Claude Sonnet 5.519.07
- #3Grok 4.518.99
- #4Claude Opus 5.518.73
- #5GPT-5.6 Sol17.93
View full science ranking →Interpret financial information and solve finance-related problems.
- #1Qwen 3.7 Max40.90
- #2Qwen 3.5 397B38.82
- #3Qwen 3.6 Plus Uncensored37.06
- #4Qwen 3.7 Plus37.01
- #5Qwen 3 235B A22B Thinking 250733.14
View full finance ranking →Return information in a precise format, such as JSON or a table.
- #1Qwen 3.6 Plus Uncensored33.92
- #2Qwen 3.7 Plus33.18
- #3Qwen 3.5 397B29.68
- #4MiniMax M3 Preview29.16
- #5Qwen 3.8 Max27.47
View full structured output ranking →Interpret legal text and reason through law-related questions.
- #1Qwen 3.7 Max44.48
- #2Qwen 3.7 Plus41.28
- #3Qwen 3.6 Plus Uncensored40.91
- #4Qwen 3.5 397B38.82
- #5NVIDIA Nemotron 3 Ultra33.49
View full legal ranking →Help with everyday work such as summarizing, organizing, and drafting.
- #1Muse Spark Open16.32
- #2Claude Opus 5.515.49
- #3Qwen 3.8 Flash14.46
- #4Hy4 Preview14.02
- #5Qwen 3.8 Max13.97
View full productivity ranking →Follow the requested format, rules, and constraints of a prompt.
- #1Qwen 3.8 Max31.75
- #2Qwen 3.7 Plus30.70
- #3Qwen 3.7 Max30.06
- #4NVIDIA Nemotron 3 Ultra29.29
- #5Qwen 3.8 Flash28.28
View full instruction following ranking →Base answers on the information provided in documents or other sources.
- #1GPT-6 Astra26.22
- #2Claude Opus 4.824.63
- #3GPT-5.223.18
- #4Qwen 3.6 27B21.90
- #5Qwen 3.8 Max21.90
View full grounding ranking →Answer questions that draw on knowledge across many subjects.
- #1Gemini 4 Argon21.52
- #2NVIDIA Nemotron 3 Ultra21.49
- #3Grok 4.518.99
- #4Gemini 3.8 Flash18.18
- #5Grok 4.717.35
View full knowledge ranking →Explain ideas clearly and adapt a response to its audience.
- #1Claude Opus 4.631.97
- #2Claude Opus 4.527.33
- #3Claude Sonnet 4.627.28
- #4Claude Sonnet 4.526.67
- #5GPT-5.225.31
View full communication ranking →Reason about markets, incentives, and economic decisions.
- #1Qwen 3.7 Max28.47
- #2Qwen 3.6 Plus Uncensored27.02
- #3Qwen 3.7 Plus25.69
- #4Qwen 3.5 397B23.50
- #5Seed 2.1 Turbo22.55
View full economics ranking →Turn visual information into useful descriptions and text answers.
- #1Qwen 3.7 Plus30.94
- #2Kimi K2.527.47
- #3GLM 5V Turbo24.78
- #4Seed 2.1 Turbo22.76
- #5NVIDIA Nemotron 3.5 Super VL 120B22.73
View full image to text ranking →Reason about how parts of a technical system work together.
- #1Gemini 4 Argon19.52
- #2GPT-5.6 Sol17.71
- #3GPT-5.6 Terra17.34
- #4GLM 5.317.19
- #5MiniMax M3.1 Flash14.96
View full systems ranking →Hold a conversation and respond usefully to follow-up questions.
- #1Qwen 3.5 397B29.56
- #2Qwen 3.7 Plus28.57
- #3Qwen 3.7 Max27.41
- #4Qwen 3.6 Plus Uncensored27.37
- #5Claude Sonnet 4.526.88
View full chat ranking →Follow actions and events in a video and answer questions about them.
- #1GLM 5.3 Flash29.88
- #2Seed 2.1 Turbo26.51
- #3Qwen 3.6 27B25.65
- #4Qwen 3.7 Plus23.99
- #5Qwen 3.8 Max23.73
View full video understanding ranking →Give accurate answers and avoid unsupported claims.
- #1Qwen 3 235B A22B Instruct 250720.46
- #2Gemini 3 Flash Preview17.95
- #3GPT-6.1 Sol15.49
- #4Inkling12.19
- #5DeepSeek V4 Flash 042311.70
View full factuality ranking →Follow written instructions to create images.
- #1Qwen 3.6 Plus Uncensored23.29
- #2Qwen 3.6 35B A3B19.80
- #3Qwen 3.6 27B15.56
- #4Qwen 3.5 35B A3B13.04
View full text to image ranking →Develop original ideas, stories, and approaches to open-ended tasks.
- #1Qwen 3 235B A22B Thinking 250727.75
- #2Qwen 3 235B A22B Instruct 250724.67
- #3NVIDIA Nemotron 3 Nano 30B13.39
View full creativity ranking →Draft, edit, and improve text for different audiences and purposes.
- #1Qwen 3 235B A22B Thinking 250727.75
- #2Qwen 3 235B A22B Instruct 250724.67
- #3NVIDIA Nemotron 3 Nano 30B13.39
View full writing ranking →Understand speech, sounds, and information in audio recordings.
- #1Inkling12.19
View full audio understanding ranking →