Mercury 2.5, Inception’s proprietary text-only reasoning model, was evaluated on Artificial Analysis’ Intelligence Index, a composite benchmark covering reasoning, knowledge, mathematics, and coding. The evaluation generated 35M output tokens, substantially below the 85M median for similarly priced reasoning models. It scored 12, at the stated comparison median, while supporting a 260k-token context window. Inference was measured through Inception’s API at 780.8 tokens per second, compared with a 110.3 t/s median for the relevant price tier; time to first token was 2.91 seconds versus a 2.22-second median. Mercury 2.5 accepts and returns text only, lacks image input and multimodal support, and is available through one API provider. Pricing is $0.25 per million input tokens and $0.75 per million output tokens, equivalent to a blended $0.14 per million tokens under the stated 7:2:1 cache-hit/input/output weighting. Artificial Analysis estimated an evaluation cost of $0.06 per Intelligence Index task. Released September 8, 2026, the model uses extended reasoning internally, but Inception has not disclosed its parameter count.