← All models

AntAngelMed

Ant Group · 102.9B parameters

AntAngelMed is an open-weight medical language model jointly developed by the Health Information Center of Zhejiang Province, Ant Healthcare, and Zhejiang Anzhen'er Medical Artificial Intelligence Technology Co., Ltd., built from InclusionAI's Ling-flash-2.0 through continual pre-training on medical corpora, supervised fine-tuning, and GRPO-based reinforcement learning. The checkpoint keeps Ling-flash-2.0's 32-layer sparse MoE architecture (256 routed experts, eight active, plus one shared expert) and contains 102,889,697,280 parameters after excluding 7,936 routing-bias buffer elements; its developers describe it as 100B total with 6.1B activated. Its released configuration is natively 32,768 tokens, which the model card says YaRN extrapolation extends to 128K (131,072 tokens), and its chat template has no thinking mode.

Benchmark scores