Granite 3.0 8B Instruct — VRAM, speed & local setup
Written by Jakub Rusinowski · Last updated
A robust 8B model optimized for enterprise tasks, coding, and RAG. Completely open license.
Granite 3.0 8B Instruct needs about 6 GB of VRAM at Q4_K_M — quantized weights plus framework overhead, before any KV cache. On Apple Silicon that figure comes out of unified memory.
VRAM and speed by quantization
Quoted against NVIDIA RTX 4090 (24 GB). Includes the KV cache at 8K context, so it reads higher than the headline figure.
| Quant | Memory | VRAM | Speed (est.) | Fit |
|---|---|---|---|---|
| Q2_K 2.63 bpw | 4.8 GB | ~151 tok/s | Fits | |
| Q3_K_M 3.41 bpw | 5.6 GB | ~131 tok/s | Fits | |
| Q4_K_M 4.83 bpw | 7 GB | ~105 tok/s | Fits | |
| Q5_K_M 5.67 bpw | 7.8 GB | ~94 tok/s | Fits | |
| Q6_K 6.56 bpw | 8.7 GB | ~85 tok/s | Fits | |
| Q8_0 8.50 bpw | 10.6 GB | ~70 tok/s | Fits | |
| F16 16.00 bpw | 18.1 GB | ~41 tok/s | Fits |
Black marker = usable memory on the NVIDIA RTX 4090 (24 GB). Estimates from the memory-bandwidth roofline on the methodology page. Granite 3.0 8B Instruct VRAM calculator →
Get Granite 3.0 8B Instruct running
The cheapest catalogued GPU that runs Granite 3.0 8B Instruct is the Intel Arc B570 (10 GB).
or compare on Vast.ai from $0.35/hr (typical low · varies)
As an Amazon Associate we earn from qualifying purchases. Cloud GPU links are referral links — we may earn a commission at no extra cost to you.
How to run Granite 3.0 8B Instruct
Install Ollama, then run:
ollama run granite3-denseSpecifications
Curated — A hand-written entry from before this catalogue recorded its sources. The figures are long-standing but their provenance is not on file.
- Parameters
- 8 Billion
- Context window
- 4,096
- Architecture
- Dense
- Provider
- IBM
- Licence
- Apache 2.0
- Practical deployment
- Laptop
- Specified at
- Q4_K_M
- System RAM
- 16 GB
- Record updated
- 2024-10-21
Commercial use permitted. No usage restrictions beyond attribution.
Quality and use cases
Scores as published by the model’s authors or an independent evaluator — quality, not throughput, and not measured by us.
Can I run Granite 3.0 8B Instruct on my GPU?
- Granite 3.0 on AMD Radeon RX 9060 XT 8GB
- Granite 3.0 on Intel Arc B570
- Granite 3.0 on Intel Arc B580
- Granite 3.0 on NVIDIA GeForce RTX 3060 (12GB)
- Granite 3.0 on NVIDIA GeForce RTX 3070
- Granite 3.0 on NVIDIA GeForce RTX 3070 Ti
- Granite 3.0 on NVIDIA GeForce RTX 3080 (10GB)
- Granite 3.0 on NVIDIA GeForce RTX 4060
- Granite 3.0 on NVIDIA GeForce RTX 4070
- Granite 3.0 on NVIDIA GeForce RTX 4070 Super
- Granite 3.0 on NVIDIA GeForce RTX 4070 Ti
- Granite 3.0 on NVIDIA GeForce RTX 5060
- Granite 3.0 on NVIDIA GeForce RTX 5060 Ti 8GB
- Granite 3.0 on NVIDIA GeForce RTX 5070
Granite 3.0 8B Instruct — frequently asked questions
How much VRAM does Granite 3.0 8B Instruct need?
About 6 GB at Q4_K_M — quantized weights plus framework overhead, before any KV cache. The cache grows with context length and is added on top; the table above folds it in. Apple Silicon counts unified memory toward the same figure.
Does Granite 3.0 8B Instruct run on an RTX 4090 (24 GB)?
Yes. Granite 3.0 8B Instruct needs about 6 GB at Q4_K_M, inside a 24 GB card, at an estimated 105 tokens/sec.
How do I run Granite 3.0 8B Instruct locally?
Install Ollama and run `ollama run granite3-dense`. That pulls the weights and starts a local OpenAI-compatible endpoint; after the download nothing leaves the machine.