Granite 4.0 Micro
Granite 4.0 Micro is an instruction model for text generation, extraction, coding and tool tasks. Its published parameter count is 3B dense. The release supports 12 languages.
Model design and training
Micro uses 40 attention layers, a hidden size of 2,560, grouped-query attention and rotary position embeddings. Unlike the H variants, it is a dense transformer without Mamba layers. It is post-trained from Granite 4.0 Micro Base.Granite 4.0 Micro model card
Practical use
Practical guidance. Micro is a useful option when the deployment engine already supports standard dense transformers. Compare its memory and latency with H-Micro on realistic input lengths. Keep retrieval quality fixed while checking whether the smaller model can cite and extract the required evidence.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | ibm-granite/granite-4.0-micro |
| Context window | 131,072 tokens Published model context; deployment settings can impose a smaller limit. Granite 4.0 Micro model card |
| License | Apache 2.0Granite 4.0 Micro model card |
| Parameters | 3B denseGranite 4.0 Micro model card |
| Architecture | Dense transformerGranite 4.0 Micro model card |
Related models
- Granite 4.1 3B
IBM · Granite 4.1
An instruction-tuned 3B text model for retrieval-based answers, extraction, coding and tool use.
Launch date API · Downloadable · Provider product - Granite 4.0 H-Micro
IBM · Granite 4.0
A 3B dense hybrid instruction model for text, code and tool tasks across 12 languages.
Launch date not confirmedAPI · Downloadable · Provider product