Skip to content
localmodel.run

Text model · Granite

Granite 3.1 2B requirements

Granite family · 2.53B params · released 2024-12-18 · 450.1K Ollama pulls. Minimum to run at Q4_K_M: Apple M1 (8GB).

LicenseApache-2.0· Commercial OK↓ 13.3K/mo♥ 56on HuggingFace
Q4_K_M
1.55 GB
Q8_0
2.69 GB
Total @ Q4 (4k)
~2.8 GB
Context
128 k

Quantization sizes

GGUF quantson disk
QuantizationSize on disk
Q2_K1.1 GB est
Q3_K_M1.2 GB est
Q4_K_M (default)1.55 GB
Q5_K_M1.8 GB est
Q6_K2.1 GB est
Q8_02.69 GB
FP165.1 GB est

Lower quant = smaller and faster, slightly lower quality. Q4_K_M is the common default.

Run it

Ollama
$ ollama run granite3.1-dense:2b
llama.cpp
$ llama-cli -hf bartowski/granite-3.1-2b-instruct-GGUF:Q4_K_M
LM Studio
$ lms get bartowski/granite-3.1-2b-instruct-GGUF

Which devices can run Granite 3.1 2B?

FAQ

How much VRAM or RAM does Granite 3.1 2B need?

At Q4_K_M, Granite 3.1 2B needs about 2.8 GB (weights ~1.55 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~3.9 GB.

Can Granite 3.1 2B run on a laptop?

Yes, Granite 3.1 2B fits on a 16 GB machine at Q4_K_M and runs on Apple Silicon or a 12 GB+ GPU comfortably.

Can I use Granite 3.1 2B commercially?

Yes. Granite 3.1 2B is licensed Apache-2.0, which permits commercial use.

IBM's enterprise-grade 2B dense SLM. 128K context (extended from 4K via progressive RoPE training on ~500B tokens). Optimized for tool-use and RAG workloads. Apache 2.0.

Sources

Memory figures are estimates. See methodology.