Open-weight model
granite-guardian-4.1-8b-smoothquant-w8a8
by Marcus Nogueira marcusnogueira/granite-guardian-4.1-8b-smoothquant-w8a8
granite-guardian-4.1-8b-smoothquant-w8a8 is an open-weight model from Marcus Nogueira. It has 8.8B parameters and a 131,072-token context. At 16-bit it needs about 21.1 GB of GPU memory, which fits on 1x MI300X from $1.85 an hour, at the lowest prices in the SAVRN Index.
SmoothQuant + GPTQ W8A8 INT8 checkpoint of ibm-granite/granite-guardian-4.1-8b at immutable revision ab01ccca5dcfb80246369a086a4a87a29198f5af.
Runs On
What it takes to serve granite-guardian-4.1-8b-smoothquant-w8a8 (8.8B parameters): the memory its weights need at each precision, and the cheapest way to rent enough data-center GPUs to hold them.
| Precision | Weights | Memory needed | Cheapest setup | Per hour | Also fits |
|---|---|---|---|---|---|
| 16-bit | 17.6 GB | 21.1 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 8-bit | 8.8 GB | 10.5 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
| 4-bit | 4.4 GB | 5.3 GB | 1x MI300X (192 GB) Vultr |
$1.85 | 1x H100 $1.99 · 1x MI325X $2.00 |
Memory is the weights at that precision plus 20% for the runtime and a short context; a long context needs more. Prices are the lowest on-demand hourly rates in the SAVRN Index, read Oct 19, 2026.
Model Card
SmoothQuant + GPTQ W8A8 INT8 checkpoint of ibm-granite/granite-guardian-4.1-8b at immutable revision ab01ccca5dcfb80246369a086a4a87a29198f5af. - Smoothing strength (alpha): 0.5 The resolved base revision, full quantization configuration, package versions, architecture mapping inventory, tensor inventory, and file hashes are included in the checkpoint metadata. vLLM smoke-test evidence is added before publication. This checkpoint passed four fixed English/Portuguese safe/unsafe smoke cases in BF16 with vLLM 0.31.0 on an NVIDIA A40. vLLM loaded the serialized compressed-tensors W8A8 configuration directly; no on-the-fly quantization was used. BF16 is the selected runtime compute dtype for the…
Excerpt from the card by Marcus Nogueira.
Configuration
- Architecture
- GraniteForCausalLM
- Context length (tokens)
- 131,072
- Layers
- 40
- Hidden size
- 4,096
- Feed-forward size
- 12,800
- Attention heads
- 32
- Key/value heads
- 8
- Vocabulary size
- 100,352
- RoPE base
- 10,000,000
- Model type
- granite
- Quantization
- compressed-tensors
Identity and Version
- Repository
- marcusnogueira/granite-guardian-4.1-8b-smoothquant-w8a8
- Publisher
- Marcus Nogueira
- Task
- Not stated by the source
- Modality
- Other
- Library
- Not stated by the source
- Parameters
- 8.8B parameters
- Languages
- en, pt
- Revision
- d7b5761947c2eb8a9fe0d06a6c68badd34ca17e0
- First published
- 2026-10-15
- Last updated
- 2026-10-15
Files and Weights
21 files, 9.6 GB in total. The weights are 3 files totalling 9.6 GB in safetensors.
Every file
| File | Type | Size | SHA-256 |
|---|---|---|---|
| model-00001-of-00003.safetensors | Weights | 4.0 GB | fdf019c54a32 |
| model-00002-of-00003.safetensors | Weights | 4.0 GB | 93ab5ba3ec25 |
| model-00003-of-00003.safetensors | Weights | 1.7 GB | a71dec629691 |
| calibration_manifest.json | Configuration | 309 B | — |
| config.json | Configuration | 2.0 KB | — |
| generation_config.json | Configuration | 147 B | — |
| model.safetensors.index.json | Configuration | 54.3 KB | — |
| quantization_manifest.json | Configuration | 6.5 KB | — |
| recipe.yaml | Configuration | 483 B | — |
| runtime_environment.json | Configuration | 584 B | — |
| special_tokens_map.json | Configuration | 579 B | — |
| verification.json | Configuration | 2.8 KB | — |
| vllm_smoke_bfloat16.json | Configuration | 6.0 KB | — |
| README.md | Documentation | 1.9 KB | — |
| SHA256SUMS | Other | 1.7 KB | — |
| chat_template.jinja | Other | 6.1 KB | — |
| .gitattributes | Repository | 1.5 KB | — |
| merges.txt | Tokenizer | 916.6 KB | — |
| tokenizer.json | Tokenizer | 7.2 MB | — |
| tokenizer_config.json | Tokenizer | 17.7 KB | — |
| vocab.json | Tokenizer | 1.6 MB | — |
License and Download
- License
- Not stated by the source
- Access
- Open weights, no gate
- Download size
- 9.6 GB
Released by Marcus Nogueira through its official repository on Hugging Face.
Built From
- Derived from ibm-granite/granite-guardian-4.1-8b
- Quantized from ibm-granite/granite-guardian-4.1-8b
Memory Requirements
| Precision | Weights in memory |
|---|---|
| As published | 9.6 GB |
| 16-bit | 17.6 GB |
| 8-bit | 8.8 GB |
| 4-bit | 4.4 GB |
Weights only, from the published parameter count; the key-value cache and runtime add to this.
Questions About granite-guardian-4.1-8b-smoothquant-w8a8
How much GPU memory does granite-guardian-4.1-8b-smoothquant-w8a8 need?
About 21.1 GB at 16-bit and 5.3 GB at 4-bit: the weights (8.8B parameters) plus a working margin. A long context needs more.
What is the cheapest GPU to run granite-guardian-4.1-8b-smoothquant-w8a8 on?
At 16-bit, 1x MI300X from $1.85 an hour; at 4-bit, 1x MI300X from $1.85 an hour, at the lowest on-demand prices the SAVRN Index lists.
What is granite-guardian-4.1-8b-smoothquant-w8a8's context length?
131,072 tokens, from the maximum position embeddings in its published configuration.