Inception: Mercury 2.5
inception's efficiency model. Context window: 260K tokens.
Scores by test
Methodology →What you need to know
Inception: Mercury 2.5 is a high-precision tool for structured data and multilingual tasks, operating at a very low price point with a blended cost of $0.122/MTok. Its primary value lies in its reliability for deterministic outputs and faithfulness, both of which earned perfect 5/5 internal scores. These capabilities, combined with a 260K context window, make it an efficient choice for extracting structured information from large, diverse datasets.
Despite its strengths in formatting and accuracy, the model ranks low overall (#136 of 164) due to significant failures in safety calibration. With a 1/5 score in this area, the model frequently over-refuses benign requests, which may disrupt workflows that require a nuanced understanding of safety boundaries. It also shows moderate limitations in tabular data processing and constrained rewriting.
Use this model if you need a low-cost, high-fidelity solution for multilingual structured output or data extraction where strict adherence to source material is required. Skip this model if your application requires flexible safety filtering or frequent manipulation of tabular data.
Strengths — Top 3
Relative weaknesses — Bottom 3
Similar models