Intern-S1-Pro: Scientific Multimodal Foundation Model at Trillion Scale

Benchmark Model Rank Results
agents-on-benchIntern-S1-Pro–Accuracy: 80.9
agents-on-gaia-text-only-103-validation-tasksIntern-S1-Pro–Accuracy: 77.4
code-generation-on-livecodebench-v6Intern-S1-Pro–Pass@1: 74.3
computer-use-agents-on-screenspot-v2Intern-S1-Pro–Accuracy: 93.6
instruction-following-on-ifbenchIntern-S1-Pro–Accuracy: 71.2
mathematical-reasoning-on-aime-2025Intern-S1-Pro–Accuracy: 93.1
multi-task-language-understanding-on-mmlu-proIntern-S1-Pro–Accuracy: 86.6
multimodal-reasoning-on-mmmu-proIntern-S1-Pro–Accuracy: 72.8
optical-character-recognition-ocr-on-ocrbench-v2Intern-S1-Pro–English Score: 60.1Chinese Score: 60.6
visual-grounding-on-refcoco-refcoco-refcocog-overallIntern-S1-Pro–Accuracy: 91.9