granite-4.0-h-small
Where it runs 8K context · ≥ 5 t/s floor
| Setup | Best plan | Memory | Generate |
|---|---|---|---|
| 8 GB GPU + 32 GB RAM | Q4_K_Mllama.cpp CUDAGPU + CPU layers · 14/40 layers on GPU | Fits* | ~4.8–6.4 t/s Estimated extrapolated |
| 12 GB GPU + 32 GB RAM | Q4_K_Mllama.cpp CUDAGPU + CPU experts · 21 expert layers on CPU | Fits* | ~3.1–4.9 t/s Estimated extrapolated |
| 16 GB GPU + 64 GB RAM | Q4_K_Mllama.cpp CUDAGPU + CPU experts · 10 expert layers on CPU | Fits* | ~10–17 t/s Estimated extrapolated |
| 24 GB GPU + 64 GB RAM | Q5_K_Mllama.cpp CUDAFull GPU | Fits* | ~30–87 t/s Estimated extrapolated |
| 32 GB GPU + 128 GB RAM | Q5_K_Mllama.cpp CUDAFull GPU | Fits* | ~54–155 t/s Estimated extrapolated |
| 2× 24 GB GPUs + 128 GB RAM | Q8_0llama.cpp CUDALayer split across GPUs | Fits* | ~19–54 t/s Estimated extrapolated |
| 96 GB workstation GPU + 256 GB RAM | FP16llama.cpp CUDAFull GPU | Fits* | ~19–55 t/s Estimated extrapolated |
| Mac, 24 GB | Does not fit | ||
| Mac, 48 GB | Q8_0llama.cpp MetalUnified memory | Fits* | ~5.5–16 t/s Estimated extrapolated |
| Mac, 128 GB | FP16llama.cpp MetalUnified memory | Fits* | ~5.8–17 t/s Estimated extrapolated |
| Mac, 512 GB | FP16llama.cpp MetalUnified memory | Fits* | ~8.7–25 t/s Estimated extrapolated |
| DGX Spark 128 GB | FP16llama.cpp CUDAUnified memory | Fits* | ~4.6–13 t/s Estimated |
| Ryzen AI Max+ 395, 128 GB | FP16llama.cpp VulkanUnified memory | Fits* | ~2.7–7.9 t/s Estimated extrapolated |
| CPU only, 64 GB DDR5-5600 dual channel | Q4_K_Mllama.cpp CPUCPU only | Fits* | ~3.8–4.5 t/s Estimated extrapolated |
Same plan engine as the planner, with its default reserves and RAM. How plans are chosen.
Files exact Hugging Face bytes
| Artifact | Size | Files | Revision | Source |
|---|---|---|---|---|
| FP16 ibm-granite/granite-4.0-h-small-GGUF | 60.02 GiB | 1 | 6522095069 | HF files |
| Q8_0 ibm-granite/granite-4.0-h-small-GGUF | 31.91 GiB | 1 | 6522095069 | HF files |
| Q5_K_M ibm-granite/granite-4.0-h-small-GGUF | 21.30 GiB | 1 | 6522095069 | HF files |
| Q4_K_M ibm-granite/granite-4.0-h-small-GGUF | 18.14 GiB | 1 | 6522095069 | HF files |
| Native safetensors BF16 ibm-granite/granite-4.0-h-small | 59.99 GiB | 14 | b8c0982bab | HF files |
Not found in the representative quant search (not proven unavailable): MLX 8-bit, MLX 4-bit, EXL3, AWQ, GPTQ. Complete selected file sets; projectors, MTP/draft and lookup files are listed separately and budgeted only by plans that use them. File bytes are not allocated memory.
Quality
| AA evaluated mode | Index | LiveCodeBench | Terminal-Bench 2.1 | GPQA | HLE | Source |
|---|---|---|---|---|---|---|
| Granite 4.0 H Small AA estimate | 6.0 | 11.3% | — | 41.6% | 3.8% | source |
Arena text rating
Unknown
Publisher-reported benchmarks
None extracted from the model card.
AA and Arena describe the hosted model in the stated mode, not each quantized file; scores are never inherited by base, pruned, abliterated or quantized variants and never averaged together.
How to run it
No model-specific recipe in the research. The planner binds the backend’s documented command template (llama.cpp, Ollama, LM Studio, MLX, vLLM) to your plan: open a plan’s details.
Recipes are source examples (settings are the publisher’s or author’s, not measured optima) and were not executed by LocalFit. A downloadable file does not prove your backend build supports the architecture.
Reported speeds 0
No reported measurement yet. Submit one.