// IBM Granite · Text LLM
granite-4.2-8b: VRAM & local deployment
8.79B parameters. About 7.7 GB at 4-bit and 21.2 GB at BF16 with one 8K-token request.
released 2026-08-07licence: Commercial OKmemory figures are estimates
01Specs
| Publisher | IBM · ibm-granite |
|---|---|
| Family / series | IBM Granite · granite-4.2 |
| Type | Text LLM (text-generation) |
| Parameters | 8.79B Source: counted from official safetensors weights |
| Context length | 128K tokens (131,072)max_position_embeddings in config.json |
| Architecture | 40 layers · GQA |
| Released weights | BF16 · 17.6 GBdownload |
| Licence | Apache-2.0 · Commercial OKPermissive licence (Apache-2.0 / MIT / CC-BY / OpenMDW): commercial use allowed. |
| Released | 2026-08-07 (Hugging Face repo created) |
02How much VRAM it needs (estimate)
Total = weights + 10% + 0.5GB overhead + KV cache for one request (FP16 KV). More concurrent users multiply the KV column.
| Precision | Weights | Total @ 8K | Total @ 32K | Total @ 128K | Smallest GPU | Mac |
|---|---|---|---|---|---|---|
| BF16 / FP16 (original) | 17.6 GB | 21.2 GB | 25.2 GB | 41.3 GB | RTX 4090 24GB | Mac 36GB |
| FP8 / INT8 / Q8_0 | 9.3 GB | 12.1 GB | 16.1 GB | 32.2 GB | 16GB GPU | Mac 24GB |
| 4-bit (Q4_K_M / AWQ / GPTQ-Int4) | 5.3 GB | 7.7 GB | 11.7 GB | 27.8 GB | 12GB GPU | Mac 16GB |
KV cache per request: 8K 1.3 GB · 32K 5.4 GB · 128K 21.5 GB
Bits per weight: BF16 16, FP8 8.5, 4-bit 4.85. “Fits” = total ≤ 90% of memory; Mac GPU can use ~2/3 of unified memory (≤36GB) or ~3/4 (larger). For MoE, all experts must be in memory, so size by total parameters. Method →
03How to run it
Only runtimes the publisher or the runtime’s official library names for this model.
- vLLM — mentioned in the official model card
- SGLang — mentioned in the official model card
- Transformers — mentioned in the official model card
- llama.cpp — official GGUF weights
- MLX — official MLX weights
- Ollama — in the official Ollama library · granite4.2:8b
ollama run granite4.2:8b04Official repositories & quantised builds
8 official repo(s) from ibm-granite.
- granite-4.2-8b-GGUF GGUF
- granite-4.2-8b main
- granite-4.2-8b-fp8 FP8
- granite-4.2-8b-nvfp4 NVFP4
- granite-4.2-8b-mxfp4 MXFP4
- granite-4.2-8b-q4-mlx MLX
- granite-4.2-8b-q8-mlx MLX
- granite-4.2-8b-bf16-mlx BF16/MLX
05Other IBM Granite models
All 67 →| Model | Params | 4-bit total | Released |
|---|---|---|---|
| granite-4.2-3b · same series | 3.66B | 3.6 GB | 2026-08-07 |
| granite-4.2-30b · same series | 29.3B | 22.2 GB | 2026-08-07 |
| granite-switch-4.1-8b | 8.78B | 7.8 GB | 2026-10-14 |
| granite-switch-4.1-30b | 30B | 22.7 GB | 2026-10-14 |
| granite-speech-5.0-2b | 2.31B | 2.0 GB | 2026-10-07 |
| ensemble-forecasting-with-ibm-granite-time-series | — | — | 2026-10-01 |
| granite-speech-5.0-470m-turboctc-nc | 473M | 0.8 GB | 2026-08-17 |
| granite-speech-5.0-470m-turboctc | 473M | 0.8 GB | 2026-08-04 |
| granite-embedding-97m-multilingual-r2 | 97M | 0.6 GB | 2026-04-20 |
| granite-embedding-311m-multilingual-r2 | 312M | 0.7 GB | 2026-04-20 |
06Sources
Data checked 2026-10-20. Parameters, context and licence are copied from the official repository; memory figures are our estimates.