Minimax: Beginning to merge M3 and H3 into one single multimodal model
Yeyi Yun of Minimax said that multimodality is the ultimate direction for foundation models, as the Chinese AI lab plans to merge its language model (M3) and multimodal model (H3) into one unified system. Yun also noted that Minimax has designed its models to be chip-agnostic, allowing them to run on both Western and domestic chips. She further addressed accusations from Anthropic regarding model distillation.
TL;DR
- Minimax is merging its language model (M3) and multimodal model (H3) into one unified system.
- Multimodality is considered the ultimate direction for foundation models.
- Minimax's models are designed to be chip-agnostic, working with both Western and domestic chips.
- The company addressed accusations from Anthropic regarding model distillation.