Granite 4.0 H-1B
Granite 4.0 H-1B is a compact instruction-tuned text model. IBM reports 1.5B parameters for this checkpoint. It belongs to the Nano releases and is adapted from its matching base model.
Model design and training
The hybrid architecture combines Mamba-2 with attention and uses no positional embeddings. Post-training includes supervised fine-tuning, reinforcement learning and model merging. This checkpoint generates text; it does not itself contain the speech components used by other Granite products.Granite 4.0 H-1B model card
Practical use
Practical guidance. H-1B offers another point between tiny local models and the 3B releases. Measure it with the runtime you will actually ship. Compare both throughput and the rate of valid structured answers, since a faster response that needs repair can increase total task time.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | ibm-granite/granite-4.0-h-1b |
| Context window | 131,072 tokens Published model context; deployment settings can impose a smaller limit. Granite 4.0 H-1B model card |
| License | Apache 2.0Granite 4.0 H-1B model card |
| Parameters | 1.5BGranite 4.0 H-1B model card |
| Architecture | Four attention layers and 36 Mamba-2 layers; hidden size 1,536.Granite 4.0 H-1B model card |
Related models
- Granite 4.0 1B
IBM ยท Granite 4.0 Nano
A 1.6B-parameter instruction model in IBM's compact Granite Nano text family.
Launch date not confirmedDownloadable