IBM: Granite 4.0 Micro
ibm-granite's efficiency model. Context window: 131K tokens.
Scores by test
Methodology →What you need to know
IBM Granite 4.0 Micro is a low-cost utility model optimized for high-volume, structured tasks. Its primary differentiator is a 131K context window paired with a perfect 5/5 internal score for long-context processing, making it capable of handling large datasets that typically require much more expensive models.
The model is priced aggressively, with a blended cost of $0.088/MTok. While it ranks low overall (#103 of 105), it delivers high value for specific narrow applications. It performs reliably in structured output, persona consistency, and multilingual tasks, all scoring 4/5.
Performance drops significantly in cognitive complexity. It struggles with strategic analysis, creative problem solving, and faithfulness, each scoring 2/5 or lower. Most notably, its safety calibration is a critical weakness at 1/5, requiring developers to implement rigorous external guardrails.
Use this model if you need a budget-friendly solution for long-document processing, basic classification, or maintaining a specific persona. Skip this model if your use case requires high factual accuracy, complex reasoning, or built-in safety filters.
Strengths — Top 3
Relative weaknesses — Bottom 3
Similar models