huggingface.co via Reddit

InternLM's 321B Sage S2.1 Flash leads MLE-bench Lite at 90.3

TL;DR

  • Shanghai Artificial Intelligence Laboratory released Sage S2.1 Flash, a 321B-parameter scientific multimodal model, on Hugging Face under MIT license.
  • The model card reports 90.3 on MLE-bench Lite, ahead of DeepSeek-V4.1-Flash-1017 at 89.2, and first place on four of fifteen benchmarks.
  • Sage SciPerception supports more than ten scientific input types, including molecular graphs, crystal structures, and protein sequences and 3D structures.

Shanghai Artificial Intelligence Laboratory has posted Sage S2.1 Flash to Hugging Face, a 321B-parameter scientific multimodal model released under MIT license. BF16 weights ship alongside an FP8 variant for lower-memory deployment.

The model card frames the system as a research assistant rather than a chat model. "Sage S2.1 Flash is a scientific multimodal model from Shanghai Artificial Intelligence Laboratory designed for research tasks that require multiple rounds of analysis and verification," it says. "It combines scientific data understanding with information retrieval, programming, and tool use. It can use specialized software to perform analyses, check results against intermediate outputs and visual feedback, and adjust subsequent steps accordingly."

The scientific side rides on a representation stack called Sage SciPerception. The card says "the model supports more than ten types of scientific input, including molecular graphs, crystal structures, and protein sequences and 3D structures." Features from each input type get encoded and projected into a shared space alongside natural language, with downstream tasks listed as prediction, classification, site localization, quantitative analysis, and structure inference.

On benchmarks, InternLM reports first place on four of fifteen. "Sage S2.1 Flash placed first on MLE-bench Lite with 90.3, ahead of DeepSeek-V4.1-Flash-1017 at 89.2," the card states, and it "placed first on OfficeQA Pro, Harvey LAB, and SpreadsheetBench 2, scoring 72.2, 25.0, and 37.1, respectively." Terminal-Bench 2.1 lands at 87.9 and DeepSWE at 67.9. "Across the 15 benchmarks and six models shown in the chart, Sage S2.1 Flash placed first on four benchmarks and second on seven."

The card publishes no accompanying paper, no training-data description, and no release date; the numbers are InternLM's own. It lands in a steady run of large open-source releases we have been tracking this quarter.