On November 22 Ant Group's Ant Ling released LingSante-3.1-flash on X, an open-weight medical model built on Ling-3.1-flash with 560B total parameters, 25B active per token and a 1M-token context window, published on Hugging Face under the MIT license. The vendor reports HealthBench Professional 65.72, HLE (text-only, with tools) 72.22 and BrowseComp (with context) 94.10, ahead of GPT-6 Astra, Claude Opus 5.5 and Gemini 4 Argon on all three; the family also includes the 124B multimodal LingSante-3.1-air-VL.
Key Takeaways
- ✓Source: the official Ant Ling X account (@AntLingAGI); weights are on Hugging Face and ModelScope under the MIT license
- ✓Specs: hybrid-linear MoE (63 KDA + 9 Gated MLA layers), 560B total and 25B active parameters, 512 routed experts with 8 active per token, 1M context
- ✓Vendor scores: HealthBench Professional 65.72, HLE text-only with tools 72.22, BrowseComp with context 94.10, plus PhysicianBench 57 and EHR-Complex 74.40; all self-reported with no third-party replication yet
- ✓Deployment: the recommended path is the SGLang day-0 image (8x H200 with TP8+EP8, or 4x GB300); vLLM is also supported and an MXFP4 quantized build is available
- ✓Limits: the vendor says it is for research and evaluation only, is not a medical device and must not be the sole basis for diagnosis, prescriptions or treatment

Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Background
Real medical work involves incomplete information, varied patients and strict compliance, so general models that answer questions well often fail at evidence checking and multi-step clinical workflows. On November 22 Ant Group's Ant Ling announced on X LingSante-3.1-flash, pitched as moving from medical QA to real healthcare tasks, with weights on Hugging Face under the MIT license.
How it works
The model keeps the hybrid-linear MoE architecture of Ling-3.1-flash: 560B total parameters, 25B active per token, 63 KDA plus 9 Gated MLA layers and a 1M-token context. General and medical data are mixed throughout training, reinforcement learning sharpens medical reasoning, tool use and search, and multi-teacher on-policy distillation merges the skills into one model. The family also includes the 124B multimodal LingSante-3.1-air-VL.
Numbers that matter
The vendor reports HealthBench Professional 65.72, HLE (text-only, with tools) 72.22 and BrowseComp (with context) 94.10, ahead of GPT-6 Astra, Claude Opus 5.5 and Gemini 4 Argon on all three, plus PhysicianBench 57 and EHR-Complex 74.40 on long-horizon workflows. These are self-reported; third-party replication has not been announced.
What it means
The recommended deployment is the SGLang day-0 image on 8x H200 or 4x GB300, with vLLM also supported and an MXFP4 quantized build available. Ant Ling stresses that it is for research and evaluation only, is not a medical device and must not be used on its own for diagnosis or prescriptions.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.