Mercury 2.5 Prioritizes Speed Over Intelligence in New Benchmarks
Inception releases Mercury 2.5, a proprietary AI model that prioritizes speed over raw intelligence. The model generates an impressive 770.4 output tokens per second, ranking it second out of 175 models tested by Artificial Analysis. This makes it one of the fastest options available for applications where response time matters most.
On the intelligence front, Mercury 2.5 scores 12 on the Artificial Analysis Intelligence Index, placing it slightly below the median of 13 among comparable models. The model remains fairly concise, producing 35M tokens during evaluation compared to a median of 85M. It supports text input and output, includes reasoning capabilities, and features a 260k-token context window that holds roughly 390 pages of text.
Pricing for Mercury 2.5 lands at the middle of the market, with input tokens costing $0.25 per million and output tokens at $0.75 per million. A generous 90% cache discount brings cached input costs down to $0.06 per million tokens. On average, running a single Intelligence Index task costs $0.06, making the model a cost-effective choice for teams that need rapid generation without premium intelligence performance.