Granite-4.0-H-Tiny
Granite-4.0-H-Tiny is an open-weights language model from IBM, released October 2, 2025. It has 6.9B parameters (mixture-of-experts, 1.5B active). The context window is 128K tokens.
We’re launching Granite 4, the next generation of IBM language models. Granite 4.0 features a new hybrid Mamba/transformer architecture that greatly reduces memory requirements without sacrificing performance.
Check it against your own PC →
Can you run Granite-4.0-H-Tiny locally?
The smallest common GPU it runs on is 6 GB of VRAM (with 32 GB of RAM): Q4_K_M, fully on GPU, roughly 110 tokens per second. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Runs on GPU | Q4_K_M (very good) | 5.0 GB | – | ~110 |
| 8 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 6.4 GB | – | ~120 |
| 12 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 6.4 GB | – | ~150 |
| 16 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 6.4 GB | – | ~200 |
| 24 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 6.4 GB | – | ~320 |
| 32 GB VRAM | Runs on GPU | Q6_K (≈ lossless) | 6.4 GB | – | ~470 |
Ready-made quantizations on Hugging Face: ibm-granite/granite-4.0-h-tiny-GGUF · unsloth/granite-4.0-h-tiny-GGUF · lmstudio-community/granite-4.0-h-tiny-GGUF · bartowski/ibm-granite_granite-4.0-h-tiny-GGUF · mradermacher/granite-4.0-h-tiny-i1-GGUF · taronaeo/Granite-4.0-H-Tiny-BE-GGUF
Benchmark results
Epoch AI has not published results for Granite-4.0-H-Tiny yet. New models are usually evaluated within days to weeks.
Links
Announcement · Hugging Face: ibm-granite/granite-4.0-h-tiny
More from IBM
granite switch 4.1 8b · granite switch 4.1 30b · Granite 4.1 30B · Granite 4.1 8B · Granite-4.0-H-Micro · Granite-4.0-H-Small · Granite-Docling