FitMyLLM
Mistral AI/Dense

Mistral AIMinistral 3 3B

Ministral 3 3B — Mistral's small dense instruct model with a 256K context, aimed at edge and on-device deployment.

chatcodingmultilingualtool_use
3.85B
Parameters
256K
Context length
14
Benchmarks
6
Quantizations
392K
HF downloads
Architecture
Dense
Released
2025-12-02
Layers
26
KV Heads
8
Head Dim
128
Family
mistral

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
Q4_K_M4.89
4.1 GB
2.8 + 1.2 KV
good
Q5_K_S5.57
4.4 GB
3.2 + 1.2 KV
good
Q5_K_M5.7
4.5 GB
3.2 + 1.2 KV
good
Q6_K6.56
4.9 GB
3.6 + 1.2 KV
excellent
Q8_08.5
5.8 GB
4.6 + 1.2 KV
lossless
FP1616
9.4 GB
8.2 + 1.2 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 Ministral 3 3B now

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

Community Ratings

Loading ratings...

Benchmarks (14)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$ollama run ministral-3:3b-instruct-2512-q4_K_M

Downloads and runs automatically. Add --verbose for speed stats.

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

GPUs that can run this model

At Q4_K_M quantization. Sorted by minimum VRAM.

NVIDIA Tesla C2050
3 GB VRAM • 144 GB/s
NVIDIA
NVIDIA Tesla M2050
3 GB VRAM • 148 GB/s
NVIDIA
NVIDIA Tesla S2050
3 GB VRAM • 148 GB/s
NVIDIA

Find the best GPU for Ministral 3 3B

Build Hardware for Ministral 3 3B
Spec sheet

Ministral 3 3B — 3.85B Dense.

Specifications
Parameters
3.85B
Architecture
Dense Transformer
Context Length
256K tokens
Capabilities
chat, coding, multilingual, tool_use
Release Date
2025-12-02
Provider
Mistral AI
Family
mistral
VRAM requirements
QuantBPWVRAMQuality
Q4_K_M4.892.8 GB94%
Q5_K_S5.573.2 GB96%
Q5_K_M5.73.2 GB96%
Q6_K6.563.6 GB97%
Q8_08.54.6 GB100%
FP16168.2 GB100%
Benchmark scores
MMLU-PRO52.4
GPQA Diamond35.8
LiveCodeBench24.7
AIME22.0
HLE5.4
AA Intelligence7.1
AA Coding4.8
AA Math22.0
IFBench26.8
τ²-Bench24.9
SciCode14.4
AA Long Context16.0
Terminal-Bench 2.10.0
Terminal-Bench 40.0
Run command

Run Ministral 3 3B locally with Ollama — needs 2.8 GB VRAM at Q4_K_M:

$ollama run ministral-3:3b
Compatible GPUs
30 @ Q4_K_M