Sage-A1.5
An open-source large model with a mixture-of-experts architecture from Shanghai AI Laboratory. It has 35B total parameters and 3B activated parameters, is positioned for intelligence capabilities in agent scenarios, and supports local deployment.
Timeline (last 90 days)
Shanghai AI Laboratory open-sourced the Sage-A1.5 model.
Sage-A1.5 uses a MoE architecture, with a total parameter count of 35B and about 3B parameters activated per inference.
vLLM announced support for Sage-A1.5 on the day the model was released.
Developers can start an inference service for Sage-A1.5 with the command `vllm serve InternScience/Sage-A1.5`.
Sage-A1.5 is positioned as "agentic intelligence", targeting agent scenarios.
All Facts (5)
Sage-A1.5 uses a MoE architecture, with a total parameter count of 35B and about 3B parameters activated per inference.
50%UnverifiedDevelopers can start an inference service for Sage-A1.5 with the command `vllm serve InternScience/Sage-A1.5`.
50%UnverifiedSage-A1.5 is positioned as "agentic intelligence", targeting agent scenarios.
50%UnverifiedShanghai AI Laboratory open-sourced the Sage-A1.5 model.
50%UnverifiedvLLM announced support for Sage-A1.5 on the day the model was released.
50%