Inception's Mercury 3 scales diffusion language models to a million-token context
Inception says Mercury 3, its diffusion LLM, is more than 40 times larger than Mercury 1. It generates 1,068 tokens a second on Nvidia GPUs and has a 1M-token context window. It costs $0.20 per million input tokens and $0.70 per million output, half that at launch. The claims that it beats speed-tier models from OpenAI and Anthropic are untested.
Inception



