Skip to content
localmodel.run

Text model · SmolLM2

SmolLM2 1.7B requirements

SmolLM2 family · 1.7B params · released Nov 2024 · 2.6M Ollama pulls · LMArena Elo 1114. Minimum to run at Q4_K_M: Apple M1 (8GB).

LicenseApache-2.0· Commercial OK↓ 122.1K/mo♥ 733on HuggingFace
Q4_K_M
1.06 GB
Q8_0
1.82 GB
Total @ Q4 (4k)
~2.2 GB
Context
8 k

Quantization sizes

GGUF quantson disk
QuantizationSize on disk
Q2_K0.7 GB est
Q3_K_M0.8 GB est
Q4_K_M (default)1.06 GB
Q5_K_M1.2 GB est
Q6_K1.4 GB est
Q8_01.82 GB
FP163.4 GB est

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

Run it

Ollama
$ ollama run smollm2:1.7b
llama.cpp
$ llama-cli -hf bartowski/SmolLM2-1.7B-Instruct-GGUF:Q4_K_M
LM Studio
$ lms get bartowski/SmolLM2-1.7B-Instruct-GGUF

Which devices can run SmolLM2 1.7B?

FAQ

How much VRAM or RAM does SmolLM2 1.7B need?

At Q4_K_M, SmolLM2 1.7B needs about 2.2 GB (weights ~1.06 GB + KV cache + overhead) at a 4k context. At Q8_0 budget ~2.9 GB.

Can SmolLM2 1.7B run on a laptop?

Yes, SmolLM2 1.7B fits on a 16 GB machine at Q4_K_M and runs on Apple Silicon or a 12 GB+ GPU comfortably.

Can I use SmolLM2 1.7B commercially?

Yes. SmolLM2 1.7B is licensed Apache-2.0, which permits commercial use.

Released by HuggingFace 2024-11-01. Q4_K_M=1.06GB, Q8_0=1.82GB from bartowski/SmolLM2-1.7B-Instruct-GGUF HF repo, cross-confirmed with HuggingFaceTB official GGUF repo. Ollama shows 1.8GB. Context 8K. Efficient edge/on-device model.

Sources

Memory figures are estimates. See methodology.