Mistral AI/Mixture of Experts

Mistral Large 4

Name: Mistral Large 4
Author: Mistral AI

Mistral Large 4 — Mistral's 1T-parameter multimodal MoE with about 52B active (49B excluding embeddings), 1M context and three reasoning-effort levels, trained from scratch in Mistral's own European datacentres. The checkpoint embeds a DSpark speculative-decoding draft head.

chatcodingreasoningmultilingualvisionmathagentictool_use

1053.16B

Parameters (52B active)

1024K

Context length

Benchmarks

Quantizations

HF downloads

Architecture

MoE

Released

2026-10-29

Layers

KV Heads

Head Dim

Family

mistral

Quantization Options

Context length:

Quant	Bits	VRAM @ 16K	Quality
IQ2_XXS	2.38	314.8 GB 313.8 + 0.9 KV	low
IQ2_M	2.93	387.2 GB 386.2 + 0.9 KV	low
Q2_K	3.16	417.4 GB 416.5 + 0.9 KV	low
IQ3_XXS	3.25	429.3 GB 428.3 + 0.9 KV	low
IQ3_XS	3.5	462.2 GB 461.2 + 0.9 KV	low
Q3_K_S	3.64	480.6 GB 479.7 + 0.9 KV	low
IQ3_M	3.76	496.4 GB 495.5 + 0.9 KV	low
Q3_K_M	4	528.0 GB 527.1 + 0.9 KV	low
Q3_K_L	4.3	567.5 GB 566.6 + 0.9 KV	moderate
IQ4_XS	4.46	588.6 GB 587.6 + 0.9 KV	moderate
Q4_K_S	4.67	616.2 GB 615.3 + 0.9 KV	moderate
Q4_K_M	4.89	645.2 GB 644.2 + 0.9 KV	good
Q5_K_S	5.57	734.7 GB 733.8 + 0.9 KV	good
Q5_K_M	5.7	751.8 GB 750.9 + 0.9 KV	good
Q6_K	6.56	865.0 GB 864.1 + 0.9 KV	excellent
Q8_0	8.5	1120.4 GB 1119.5 + 0.9 KV	lossless
FP16	16	2107.8 GB 2106.8 + 0.9 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 ~315 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 4 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...

Run this model

▸Easiest way to get started·Beginners

DOCS ↗

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

$ollama run mistral:1053b-q4_K_M

Tag may need adjustment — check ollama.com/library/mistral 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 GGUF Downloads Build Hardware

Find the best GPU for Mistral Large 4

Build Hardware for Mistral Large 4

QUANT	BPW	VRAM	QUALITY
IQ2_XXS	2.38	313.8 GB	65%
IQ2_M	2.93	386.2 GB	75%
Q2_K	3.16	416.5 GB	78%
IQ3_XXS	3.25	428.3 GB	82%
IQ3_XS	3.5	461.2 GB	84%
Q3_K_S	3.64	479.7 GB	85%
IQ3_M	3.76	495.5 GB	86%
Q3_K_M	4	527.1 GB	88%
Q3_K_L	4.3	566.6 GB	90%
IQ4_XS	4.46	587.6 GB	92%
Q4_K_S	4.67	615.3 GB	93%
Q4_K_M	4.89	644.2 GB	94%
Q5_K_S	5.57	733.8 GB	96%
Q5_K_M	5.7	750.9 GB	96%
Q6_K	6.56	864.1 GB	97%
Q8_0	8.5	1119.5 GB	100%
FP16	16	2106.8 GB	100%

Mistral Large 4

Quantization Options

Community Ratings

Run this model

Auto-setup with fitmyllm CLI

Mistral Large 4 — 1053.16B MoE.