Key Info

Inception AI's Mercury 3, described as the most intelligent diffusion LLM released so far, is now live on OpenRouter with generation speeds above 1,000 tokens per second and a 1M-token context window.

Highlights

  • Generates 1,000+ tok/s, roughly 4x faster than Mercury 2.5
  • 1M context window
  • Inception reports it outperforms GPT-6 Luna and Haiku 5.5 on IFBench, GPQA Diamond, and SWE-Bench Atlas
  • Launch pricing: $0.10/M input tokens, $0.35/M output tokens (50% off)
  • Available immediately via OpenRouter