Benchmark Almanac

Shanghai AI Lab

44 models 5 model cards read

What its model cards measure

Tall ticks are Shanghai AI Lab model cards that list the benchmark in their headline table. Gray dots are cards that leave it out. Rows are the benchmarks the lab listed most, at least twice each, in the order it started using them.

MMLU-Pro GPQA Diamond IFEval MATH-500 AIME 2024 AIME 2025 Scientists' First Exam (SFE) MicroVQA XLRS-Bench MMMU MMStar MathVista MATH-Vision (MathVision) SMolInstruct Matbench MSEarth ScreenSpot-v2 SciReasoner evaluation suite Mol-Instructions Biology-Instructions MMMU-Pro

Model cards

DateModelDocumentBenchmarks
13 Sep 2026 Intern-S2-397B Intern-S2-397B model card (Hugging Face) 21
2 Feb 2026 Intern-S1-Pro Intern-S1-Pro model card (Hugging Face) 21
25 Aug 2025 InternVL3.5-241B-A28B InternVL3.5-241B-A28B model card (Hugging Face) 37
24 Jul 2025 Intern-S1 Intern-S1 model card (Hugging Face) 17
13 Jan 2025 InternLM3-8B-Instruct InternLM3-8B-Instruct model card (Hugging Face) 16

Every Shanghai AI Lab model