Sand.ai Open-Sources 114B Audio-Video Model MAGI-2 Preview
Sand.ai released and open-sourced MAGI-2 Preview, a 114B-parameter audio-video generation model. By utilizing an extreme MoE architecture that activates only about 6B parameters per token, it reduces inference costs to a tenth of the industry average.
2026-08-05 ~ 2026-08-05 · 2 related posts
- Sand.ai Open-Sources 100B-Param Video Model MAGI-2 at 1/10th Inference Cost — 量子位 · 2026-08-05
- SandAI Releases MAGI-2 Preview: 114B Audio-Video Generation Model — Nice_Amphibian_8367 · 2026-08-05