Inception has released Mercury 3, a diffusion language model capable of generating 1,068 tokens per second with a 1 million token context window. This speed is achieved through a diffusion process that refines the entire output simultaneously over several passes, unlike traditional autoregressive models. The near-instantaneous generation could enable new applications such as generating multiple candidate responses for selection, performing comprehensive pre-display checks, and enabling more complex agentic workflows. AI
IMPACT Near-instantaneous text generation could enable new AI applications like multi-candidate response selection and more complex agentic workflows.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →