Granite 4.0 H Micro
Tiny hybrid Mamba-2/Transformer model for cheap, long-context enterprise tasks.
Granite-4.0-H-Micro is a 3B-parameter model from IBM’s Granite 4.0 family, whose “H” models interleave Mamba-2 state-space layers with transformer attention to cut memory use on long inputs and concurrent sessions. Apache 2.0.
- Provider
- IBM
- Type
- Language model
- Released
- Oct 2, 2025
- Context window
- 131K tokens
- Max output
- 118K tokens
- Price
- $0.017 input / $0.11 output per 1M tokens
- Input
- text
- Output
- text
- Open weights
- Yes
- License
- Apache 2.0
Best for
- Local & on-device
- Extraction & classification
- Enterprise
- High volume / low cost
Strengths
- Hybrid Mamba-2 architecture
- Apache 2.0
- Runs on modest GPUs
More from IBM
- Granite 4.2 8B — IBM. Apache-licensed small reasoning model for governed enterprise use.