Modal announces availability of DeepSeek v4.1-Flash
Modal says the new multimodal model targets better cost efficiency for input-heavy AI agent workloads through DeepSeek’s Causal Encoder-Decoder architecture.
TLDR
Modal announced on September 11 that DeepSeek v4.1-Flash was available on its platform. According to Modal, the multimodal model’s Causal Encoder-Decoder architecture activates 16 billion parameters for input and 8 billion for output—a design intended to improve cost efficiency for input-heavy AI agent workloads.
Combined views
—
1 Source, first seen ago
