Inception releases flagship diffusion language model Mercury 3

Inception AI Inc. releases its flagship diffusion language model Mercury 3. Information published by OpenRouter on October 20 shows that Mercury 3 is now available on the platform, with a generation speed exceeding 1000 tok/s, a context window of 1M, and a speed 4 times that of Mercury 2.5; Inception claims that it surpasses GPT-6 Luna and Haiku 5.5 on IFBench, GPQA Diamond, and SWE-Bench Atlas, and during its launch period the pricing is $0.10/M tokens for input and $0.35/M tokens for output, equivalent to a 50% discount. SiliconANGLE reported on October 21 that Inception outperforms GPT-6 Luna and Claude Haiku 5.5 on 8 of its own 12 benchmarks, with speeds 4 to 8 times faster; the company says it reaches 1068 tokens per second on NVIDIA GPUs, compared with about 240 for Haiku 5.5 and about 130 for GPT-6 Luna; the context window has been increased to 1 million tokens, the parameter count was not disclosed, with only a claim that it exceeds Mercury 1 by more than 40 times, and pricing is 2 per million input tokens

AI-generated from public reports

1. Oct 20· 2 reports

   Inception releases flagship diffusion language model Mercury 3

   Source:OpenRouter