HuggingFace/Dense

SmolLM3-3B

Name: SmolLM3-3B
Author: HuggingFace

SmolLM3 is a 3B parameter language model designed to push the boundaries of small models. It supports dual mode reasoning, 6 languages and long context. SmolLM3 is a fully open model that offers strong performance at the 3B–4B scale.

chat

3.1B

Parameters

64K

Context length

Benchmarks

Quantizations

1.1M

HF downloads

Architecture

Dense

Released

2025-07-01

Layers

KV Heads

Head Dim

128

Family

smollm

Quantization Options

Context length:

Quant	Bits	VRAM @ 16K	Quality
Q4_K_M	4.89	3.2 GB 2.4 + 0.8 KV	good
Q5_K_S	5.57	3.5 GB 2.6 + 0.8 KV	good
Q5_K_M	5.7	3.5 GB 2.7 + 0.8 KV	good
Q6_K	6.56	3.9 GB 3.0 + 0.8 KV	excellent
Q8_0	8.5	4.6 GB 3.8 + 0.8 KV	lossless
FP16	16	7.5 GB 6.7 + 0.8 KV	lossless

Select your GPU above to see speed estimates and compatibility for each quantization.

Deploying for a team or in production? Size GPUs, cost & scaling in Enterprise →

▸ READY TO RUN THIS?RENT BY THE HOUR

RENT A GPU AND RUN SMOLLM3-3B NOW

Rent on Vast.ai →Or RunPod →

Spin up an A100 / H100 / 4090 in ~60s. Pay by the second. Cancel anytime.

Community Ratings

Loading ratings...

Benchmarks (7)

IFEval76.7

MATH46.1

GPQA35.7

HumanEval30.5

BBH10.9

MMLU-PRO10.7

MUSR2.8

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

curl -fsSL https://ollama.com/install.sh | sh

$ollama run smollm:3b

Tag may need adjustment — check ollama.com/library/smollm for available tags.

▸ SETUP GUIDE

Auto-setup with fitmyllm CLI

Detects your GPU, recommends the best model, downloads it, and starts chatting — zero config. Benchmarks your speed and contributes anonymous data to improve predictions.

pip install fitmyllmthen run fitmyllmLearn more

Auto-detect GPULive tok/s in chatSpeed benchmarks9 inference engines

HuggingFace Ollama Library GGUF Downloads Build Hardware

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA Quadro 5000

3 GB VRAM • 120 GB/s

NVIDIA

QUANT	BPW	VRAM	QUALITY
Q4_K_M	4.89	2.4 GB	94%
Q5_K_S	5.57	2.6 GB	96%
Q5_K_M	5.7	2.7 GB	96%
Q6_K	6.56	3.0 GB	97%
Q8_0	8.5	3.8 GB	100%
FP16	16	6.7 GB	100%