Granite 4.0 H-Tiny
Granite 4.0 H-Tiny is an instruction model for text generation, extraction, coding and tool tasks. Its published parameter count is 7B total; 1B active. The release supports 12 languages.
Model design and training
H-Tiny combines four attention layers and 36 Mamba-2 layers with a mixture-of-experts feed-forward design. It has 64 experts, with six active experts in the published architecture table. Its hidden size is 1,536.Granite 4.0 H-Tiny model card
Practical use
Practical guidance. H-Tiny may reduce per-token computation, but all expert weights still need storage. Size the deployment from total model memory as well as active parameters. Measure several simultaneous requests to see whether routing and serving overhead offset the compute benefit.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | API, Downloadable, Provider product |
| Model identifiers | ibm-granite/granite-4.0-h-tiny |
| Context window | 131,072 tokens Published model context; deployment settings can impose a smaller limit. Granite 4.0 H-Tiny model card |
| License | Apache 2.0Granite 4.0 H-Tiny model card |
| Parameters | 7B total; 1B activeGranite 4.0 H-Tiny model card |
| Architecture | Hybrid Mamba-2 and attentionGranite 4.0 H-Tiny model card |
| Hosted access | watsonx: deploy on demand.IBM watsonx.ai foundation model library |
Related models
- 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 - Granite 4.0 H-Small
IBM · Granite 4.0
A hybrid instruction model with 32B total and 9B active parameters for text, code and tool tasks.
Launch date not confirmedAPI · Downloadable · Provider product