Physix FrontierPhysix Frontier · News Briefing Card (enbrief · Sep 25, 2026)
Moore Threads MTT S5000 Completes Protenix-v2 Inference Adaptation
KEY FACTS
- Moore Threads announced on September 24 that its MTT S5000 AI compute card has completed inference adaptation for Protenix-v2.
- Protenix-v2 was developed by ByteDance's Seed team and open-sourced in April 2026.
- The model can predict the structures of proteins, DNA, RNA, and small-molecule ligand complexes end to end.
- Moore Threads says it now covers three major AI4S scenarios: macromolecule prediction, genome modeling, and medical imaging.
PHYSIX OBSERVATION
A domestic GPU adapting a domestic open-source model is a solid move toward compute self-reliance in AI4S. ByteDance's model benchmarks against AlphaFold 3, and Moore Threads fills in the inference foundation, meaning domestic pharma companies and research institutions gain another path that does not depend on overseas compute. But adaptation is only the starting point; the ecosystem toolchain and actual benchmark scores are the real test.
Source: enbrief original report ↗