Inkling Mixture-of-Experts Model with 975B Parameters and 1M Context Window
July 15, 2026
Inkling is a Mixture-of-Experts transformer featuring 975B total parameters and 41B active parameters. It was pretrained on 45 trillion multimodal tokens and supports a 1M token context window. A lighter Inkling-Small variant offers 12B active parameters for improved latency and cost-efficiency.
HOW THIS AFFECTS YOU
●
builderYou can deploy a massive-scale MoE model with 1M context for complex multimodal tasks.
●
researcherThe 45T token multimodal training recipe and MoE architecture provide a new large-scale baseline.