Introducing Claude 3.5 Sonnet

Benchmark Model Rank Results
arithmetic-reasoning-on-gsm8kClaude 3.5 Sonnet–Accuracy: 96.4
chart-question-answering-on-chartqaClaude 3.5 Sonnet–1:1 Accuracy: 90.8
code-generation-on-humanevalClaude 3.5 Sonnet–Pass@1: 92.0
code-generation-on-humanevalClaude Sonnet 3.5–Pass@1: 90.85
math-word-problem-solving-on-mathClaude 3.5 Sonnet–Accuracy: 71.1
multi-task-language-understanding-on-mmluClaude 3.5 Sonnet–Average (%): 88.7
multimodal-reasoning-on-mathvistaClaude 3.5 Sonnet–Accuracy: 67.7
multimodal-reasoning-on-mmmuClaude 3.5 Sonnet–Accuracy: 68.3
multiple-choice-question-answering-mcqa-on-gpqa-diamondClaude 3.5 Sonnet–Accuracy: 59.4
visual-question-answering-on-docvqa-testClaude 3.5 Sonnet–ANLS: 0.952