FitMyLLM
Mistral AI/Mixture of Experts

Mistral AIMistral Large 3

Mistral Large 3 — 675B MoE flagship. Frontier-class with vision and tool use.

chatcodingreasoningvisiontool_useTool Use
675B
Parameters (39B active)
256K
Context length
16
Benchmarks
17
Quantizations
1K
HF downloads
Architecture
MoE
Released
2025-12-02
Layers
88
KV Heads
8
Head Dim
128
Family
mistral

Quantization Options

Context length:
QuantBitsVRAM @ 16KQuality
IQ2_XXS2.38
205.4 GB
201.3 + 4.1 KV
low
IQ2_M2.93
251.8 GB
247.7 + 4.1 KV
low
Q2_K3.16
271.2 GB
267.1 + 4.1 KV
low
IQ3_XXS3.25
278.8 GB
274.7 + 4.1 KV
low
IQ3_XS3.5
299.9 GB
295.8 + 4.1 KV
low
Q3_K_S3.64
311.7 GB
307.6 + 4.1 KV
low
IQ3_M3.76
321.9 GB
317.7 + 4.1 KV
low
Q3_K_M4
342.1 GB
338.0 + 4.1 KV
low
Q3_K_L4.3
367.4 GB
363.3 + 4.1 KV
moderate
IQ4_XS4.46
380.9 GB
376.8 + 4.1 KV
moderate
Q4_K_S4.67
398.6 GB
394.5 + 4.1 KV
moderate
Q4_K_M4.89
417.2 GB
413.1 + 4.1 KV
good
Q5_K_S5.57
474.6 GB
470.5 + 4.1 KV
good
Q5_K_M5.7
485.6 GB
481.4 + 4.1 KV
good
Q6_K6.56
558.1 GB
554.0 + 4.1 KV
excellent
Q8_08.5
721.8 GB
717.7 + 4.1 KV
lossless
FP1616
1354.6 GB
1350.5 + 4.1 KV
lossless

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

Too big for a single GPU — plan a multi-GPU deployment
Even the lightest quant needs ~205 GB. Size GPUs, replicas, TCO and scaling for a production setup. Open in Enterprise →
Ready to run this?Rent by the hour

Rent a GPU and run Mistral Large 3 now

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

Community Ratings

Loading ratings...

Benchmarks (16)

Run this model

Easiest way to get started·Beginners
Docs ↗
curl -fsSL https://ollama.com/install.sh | sh
$# Mistral Large 3 is not in Ollama's library, and we know no GGUF build of it

Search Hugging Face for a GGUF of it (https://huggingface.co/models?sort=trending&search=Mistral-Large-3-675B-Instruct-2512+GGUF), or use the vLLM tab with the original weights.

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

Find the best GPU for Mistral Large 3

Build Hardware for Mistral Large 3

Mistral Large 3 — 675B MoE flagship. Frontier-class with vision and tool use.

Spec sheet

Mistral Large 3 — 675B MoE.

Specifications
Parameters
675B (39B active)
Architecture
Mixture of Experts
Context Length
256K tokens
Capabilities
chat, coding, reasoning, vision, tool_use
Release Date
2025-12-02
Provider
Mistral AI
Family
mistral
VRAM requirements
QuantBPWVRAMQuality
IQ2_XXS2.38201.3 GB65%
IQ2_M2.93247.7 GB75%
Q2_K3.16267.1 GB78%
IQ3_XXS3.25274.7 GB82%
IQ3_XS3.5295.8 GB84%
Q3_K_S3.64307.6 GB85%
IQ3_M3.76317.7 GB86%
Q3_K_M4338.0 GB88%
Q3_K_L4.3363.3 GB90%
IQ4_XS4.46376.8 GB92%
Q4_K_S4.67394.5 GB93%
Q4_K_M4.89413.1 GB94%
Q5_K_S5.57470.5 GB96%
Q5_K_M5.7481.4 GB96%
Q6_K6.56554.0 GB97%
Q8_08.5717.7 GB100%
FP16161350.5 GB100%
Benchmark scores
MMLU-PRO51.5
GPQA Diamond68.0
LiveCodeBench46.5
AIME38.0
HLE4.1
AA Intelligence22.8
AA Coding22.7
AA Math38.0
IFBench36.2
Terminal-Bench Hard15.9
τ²-Bench24.6
SciCode36.2
AA Long Context34.7
Terminal-Bench 2.112.0
Terminal-Bench 40.0
EQ-Bench Creative Writing v31413