IBM: Granite 4.0 Micro

IBM Granite 4.0 Micro is positioned within IBM's Granite 4 family as a compact, text-to-text model fine-tuned specifically for long-context tool calling. Its 131,000-token context window makes it well suited to workflows that need to ingest lengthy documents, multi-turn agent traces, or large retrieval-augmented prompts without aggressive truncation. The model is described as carrying roughly 3 billion parameters, placing it in the lightweight end of the Granite 4 line and making it a practical choice for teams that want Granite 4 behavior at lower compute cost than the larger siblings.

Granite 4.0 Micro is exposed alongside other Granite variants through aggregator gateways that route requests through a unified API, making it straightforward to plug into existing applications and compare against peers in the same family. Documented use cases include generating detailed text content, analyzing lengthy documents, customizing outputs through parameter tuning such as temperature and frequency penalty, and enhancing text-based applications. The combination of a very large context window with a tool-calling focus gives the model a clear fit for agent-style orchestration, document reasoning, and RAG pipelines where keeping long histories in context matters more than maximizing raw single-shot generation quality.