ModelsArchitectures & capability
Shanghai AI Lab open-sources Sage S2.1 Flash, a multimodal mixture-of-experts model for iterative scientific research
The MIT-licensed model has about 320 billion total parameters, with 18 billion active. According to a community post, it targets multi-round scientific analysis and verification and combines data understanding, retrieval, coding and tool use.
From the last30days · reddit article of November 16, 2026 and cited sources: Sage S2.1 Flash is reported to be an open-source multimodal mixture-of-experts model under the MIT license, with about 320B total parameters and 18B activated. [1]
SOURCES[1] Shanghai AI Lab launches Sage S2.1 Flash, a scientific… reddit.com
AI-assisted, not human-edited. Methodology Corrections Who owns this site
- What happened
- Sage S2.1 Flash is reported to be an open-source multimodal mixture-of-experts model under the MIT license, with about 320B total parameters and 18B activated. [1]
- Original source
- last30days · reddit article, independent perspective
- Brief by
- AT Pulse research desk AI-assisted; publisher and contact on the desk page
- When
- November 16, 2026 article published
- Covered
- Nov 17 edition
- Evidence
- 1 source across 1 site
Reach: not measured yet
Sage S2.1 Flash is reported to be an open-source multimodal mixture-of-experts model under the MIT license, with about 320B total parameters and 18B activated. [1]
The model is described as built for multi-round scientific work, combining scientific data understanding, information retrieval, coding and tool use to support analysis and verification. [1]
A permissively licensed, sparsely activated model built for repeated rounds of scientific analysis could give researchers an open alternative for agent-style scientific work.
Read the full assessment
Its usefulness depends on performance claims that have not yet been independently evaluated.
Shanghai Artificial Intelligence Laboratory has open-sourced Sage S2.1 Flash, a multimodal mixture-of-experts model built for multi-round scientific analysis and verification. The model has about 320 billion total parameters with 18 billion activated and is released under the MIT license. It combines scientific data understanding, information retrieval, coding, and tool use. Through its Sage SciP