Inception AI's new diffusion model Mercury 3 is now live on OpenRouter, generating over 1,000 tokens per second with a 1 million-token context and 4x the performance of Mercury 2.5. Inception says it beats GPT-6 Luna and Haiku 5.5 on IFBench, GPQA Diamond, and SWE-Bench Atlas, with launch pricing discounted 50% to $0.10/M input and $0.35/M output.
Inception AI's decision model Mercury Decide is now live on OpenRouter with free early access. The model takes application state and structured questions as input, and returns structured answers with probabilities attached. Inception says it is the most intelligent decision model on OpenRouter on JevBench v1.4.