Inception Labs

AI company developing diffusion-based language models that generate full token sequences simultaneously rather than autoregressively, achieving very high token throughput for low-latency applications.