Inception has announced that its Mercury 3 diffusion language model outperforms competitors GPT-6 Luna and Claude Haiku 5.5 in both speed and benchmark testing. According to The Decoder, Mercury 3 operates four to eight times faster than these rival models.
The Decoder also reports that Mercury 3 leads in eight out of 12 benchmark tests, positioning it as the most capable diffusion language model released so far, according to Inception’s claims. However, The Decoder notes that no independent verification of these results has been conducted to date.
As Japan’s markets increasingly embrace AI-driven technologies across FX, crypto, and equities trading, advancements in language models like Mercury 3 could influence algorithmic strategies and data analysis tools used locally.