Ant Medical’s Sante model is open-sourced, with its professional medical score rising to 54.29
BIBIBI
AT A GLANCE
Ant Group’s Bailing has open-sourced the Sante medical model and announced higher scores in professional medical tests.
Article
Ant Medical’s Sante model is open-sourced, with its professional medical score rising to 54.29 Dongcha Beating AI breaking news: Ant Group’s Bailing has officially open-sourced its large medical model Ling-3.0-flash-Sante, using the MIT license.
It is based on Ling-3.0-flash to strengthen its medical capabilities. It uses an MoE architecture, with a total of 124000000000, but activates only approximately 5100000000 parameters per token. Its key capabilities include medical reasoning, drug safety, evidence-based retrieval, and long-horizon medical research.
According to the latest official evaluation, Sante’s score in the HealthBench Professional professional medical test rose from 45.73 points to 54.29 points. Its score in the complex medical reasoning test rose from 65.22 points to 76.92 points, an increase of 11.7 points. In addition to answering medical questions, it also strengthens literature retrieval, tool use, and multi-agent collaboration, aiming to help researchers complete more complex medical research tasks.
Original link https://m.theblockbeats.info/flash/373401
Key points
01
Ant Group’s Bailing has officially open-sourced its large medical model Ling-3.0-flash-Sante。
02
The model uses a MIT license and is based on Ling-3.0-flash to strengthen its medical capabilities.
03
The model uses an MoE architecture, with total parameters of 124000000000, and activates approximately 5100000000 parameters per token.
04
The latest official evaluation shows that HealthBench Professional’s score rose from 45.73 points to 54.29 points.
05
The complex medical reasoning test score rose from 65.22 points to 76.92 points, an increase of 11.7 points.
06
The model strengthens literature retrieval, tool use, and multi-agent collaboration.
AI-assisted interpretation
The following is analysis, separate from reported facts. Verify important claims independently.
Sante is a large model designed for medical tasks. It not only answers medical questions but also attempts to handle medical reasoning, drug safety, literature retrieval, and medical research workflows. The reported score increases indicate higher performance on the specified tests, but do not mean that its effectiveness in real-world medical settings has been sufficiently validated.
Why it matters to readers
Open sourcing and higher scores in professional medical tests may make it easier for researchers to try and study this model.
You can understand it as an open-source artificial intelligence model with enhanced medical capabilities, but test results cannot replace a doctor’s diagnosis or independent validation.
Risks and unknowns
The current evidence does not provide an official announcement or open-source repository link from Ant Group’s Bailing.
HealthBench Professional The complete evaluation methods for the professional medical tests and complex medical reasoning tests were not explained in the evidence.
The evidence does not state whether these results underwent independent retesting.
The model’s performance in actual medical research or clinical settings; the evidence is not specified.
Related Developments
Loading event timeline…
Related concepts
MoE architecture
This term is not in the glossary yet. Browse related concepts in the glossary.
