Mistral Large 3 675B Instruct 2512 — Hardware Requirements & GPU Compatibility

ChatFunctions

Mistral-Large-3-675B-Instruct-2512 is Mistral AI's flagship instruction-tuned model, a multimodal granular Mixture-of-Experts system pairing a roughly 673-billion-parameter, about 39-billion-active-parameter language backbone with a 2.5-billion-parameter vision encoder, for around 675 billion total and 41 billion active parameters overall. It handles vision alongside text, supports dozens of languages, and is built for agentic use with native function calling and JSON output, aimed at long-document understanding, coding, and enterprise knowledge work rather than dedicated step-by-step reasoning. This is a frontier-scale model that requires a full multi-GPU server node to run, even quantized. Context length is 262,144 tokens. It is released under the Apache 2.0 license, permitting unrestricted commercial and research use. It was published in November 2025, as the third generation of Mistral's Large model line, alongside FP8, NVFP4, and BF16 weight releases.

675B params · MoE41B activeApache 2.0Nov 20252.5K downloads261 likes6.6K quant downloads

Your hardware

Verdicts below are for the reference RTX 4090. Set yours to personalise the page.

Detection reads your browser's GPU name with no data sent anywhere.

Mistral Large 3 675B Instruct 2512 runs on a 512 GB+ Mac or a multi-GPU server

Best local pick → Mac Studio (M3 Ultra, 512GB) · Q4_K_M · 454.4 GB · ~18 tok/s

Macs from 256 GB (IQ2_S · ~29 tok/s)

Score 89.6#75 of 106 open#71 of 102 that fit ≤512 GBApache 2.0

Not set up for it?

Reference GPU · RTX 4090 24 GB

Estimates, ±20% · how we estimate

llmrun Composite Score →

89.6

90% range 81.5–97.6
#75 of 106 open models
#71 of 102 that fit ≤512 GB

Benchmarks · tier chart ↓

Among 106 open modelsHow the Composite Score is built

Coding92.890% range 76–110
AgentsNot enough data yet
MathNot enough data yet
Science96.990% range 75–119
Reasoning89.090% range 80–98

Run Mistral Large 3 675B Instruct 2512

One recommended way. Every other method, OS and quant is in the step-by-step guide.Each way opens a step-by-step guide for your OS and quant.

Ways to runOther ways

each opens step-by-step

Run in cloud

Rent 7× H100 SXM · Q4_K_M

Estimated hourly price: ~$18.83/hr; est., whole server
GPUs: 7 GPUs; 560 GB of VRAM

Deployon RunPodRent H100 SXM$2.69/hr per GPU

$5+ bonus for new users

Compare GPUs →

How Much VRAM Does Mistral Large 3 675B Instruct 2512 Need?

Select a quantization level; the listed hardware adheres to it.

VRAM	+32K	+128K	File	Quality	Runs on
204.2 GB	223.2 GB	284.2 GB	224.8 GB	2-bit · i-quant extreme compression — Importance-weighted 2-bit, extreme compression — significant quality loss	Runs on Instinct MI325X 256 GB
222.8 GB	241.8 GB	302.7 GB	189.5 GB	2-bit · i-quant extra small — Importance-weighted 2-bit, extra small	Runs on Instinct MI325X 256 GB
232.0 GB	251.1 GB	312 GB	190.8 GB	2-bit · i-quant small — Importance-weighted 2-bit, small	Runs on Instinct MI325X 256 GB
250.6 GB	269.6 GB	330.6 GB	235.4 GB	2-bit · i-quant medium — Importance-weighted 2-bit, medium	Runs on Instinct MI350X 288 GB
287.7 GB	306.8 GB	367.7 GB	279.2 GB	3-bit · i-quant — Importance-weighted 3-bit	Runs on 5× H100 SXM 80 GB
306.3 GB	325.3 GB	386.2 GB	279.1 GB	3-bit · i-quant extra small — Importance-weighted 3-bit, extra small	Runs on 5× H100 SXM 80 GB
315.6 GB	334.6 GB	395.5 GB	247.1 GB	2-bit · K-quant — 2-bit quantization with K-quant improvements	Runs on 5× H100 SXM 80 GB
324.8 GB	343.9 GB	404.8 GB	291.9 GB	3-bit small — 3-bit small quantization	Runs on 5× H100 SXM 80 GB
334.1 GB	353.2 GB	414.1 GB	309.9 GB	3-bit · i-quant medium — Importance-weighted 3-bit, medium	Runs on 5× H100 SXM 80 GB
362.0 GB	381 GB	441.9 GB	321.8 GB	3-bit medium — 3-bit medium quantization	Runs on 6× H100 SXM 80 GB
371.3 GB	390.3 GB	451.2 GB	381.7 GB	4-bit · legacy — 4-bit legacy quantization	Runs on 6× H100 SXM 80 GB
380.5 GB	399.6 GB	460.5 GB	321.7 GB	3-bit large — 3-bit large quantization	Runs on 6× H100 SXM 80 GB
399.1 GB	418.1 GB	479.1 GB	360 GB	4-bit · i-quant compact — Importance-weighted 4-bit, compact	Runs on 6× H100 SXM 80 GB
417.7 GB	436.7 GB	497.6 GB	422.4 GB	4-bit · legacy+ — 4-bit legacy quantization with offset	Runs on 6× H100 SXM 80 GB
417.7 GB	436.7 GB	497.6 GB	382.7 GB	4-bit small — 4-bit small quantization	Runs on 6× H100 SXM 80 GB
417.7 GB	436.7 GB	497.6 GB	380.7 GB	4-bit · i-quant non-linear — Importance-weighted 4-bit, non-linear	Runs on 6× H100 SXM 80 GB
445.5 GB	464.5 GB	525.5 GB	407 GB	4-bit medium — 4-bit medium quantization — most popular sweet spot	Runs on 7× H100 SXM 80 GB
510.5 GB	529.5 GB	590.4 GB	464.4 GB	5-bit small — 5-bit small quantization	Runs on 8× H100 SXM 80 GB
529.0 GB	548.1 GB	609 GB	478 GB	5-bit medium — 5-bit medium quantization — good quality/size tradeoff	Runs on 8× H100 SXM 80 GB
612.6 GB	631.6 GB	692.5 GB	553.3 GB	6-bit, very good — 6-bit quantization, very good quality	Runs on 5× H200 SXM 141 GB
742.5 GB	761.5 GB	822.5 GB	715.7 GB	8-bit, near-lossless — 8-bit quantization, near-lossless	Runs on 6× H200 SXM 141 GB
1485 GB	1504 GB	1565 GB	1347 GB	16-bit · bfloat — Brain floating point 16 — preferred for training	Runs on 3 nodes of 8× H100 SXM

Selected Q4_K_M

4-bit medium quantization

454.4GBat 16K context

Weights: 443.9 GB
KV cache: 10.2 GB
Overhead: 0.3 GB

Server-class: ~454.4 GB at 16K context · about 7× H100 SXM 80 GB

Which Hardware Can Run Mistral Large 3 675B Instruct 2512?

Q4_K_M · 445.5 GB

Mistral Large 3 675B Instruct 2512 (Q4_K_M) requires 445.5 GB of VRAM to load the model weights. For comfortable inference with headroom for KV cache and system overhead, 580+ GB is recommended. No consumer GPU has enough memory. 1 device with unified memory can run Mistral Large 3 675B Instruct 2512, including Mac Studio (M3 Ultra, 512GB).

Needs server-class hardware

Mistral Large 3 675B Instruct 2512 runs on a 512 GB+ Mac or a multi-GPU server · Q4_K_M needs ~454.4 GB

Hardware that runs this model
Hardware	Quant	Memory	Speed	Context	Hardware page
Mac 256 GB · Mac Studio (M3 Ultra, 256GB)	IQ2_S	240.9 GB	~29 tok/s	32K ctx	Mac Studio →	IQ2_S · 240.9 GB · ~29 tok/s · 32K ctx
Mac 512 GB · Mac Studio (M3 Ultra, 512GB)	Q4_K_M	454.4 GB	~18 tok/s	32K ctx	Mac Studio →	Q4_K_M · 454.4 GB · ~18 tok/s · 32K ctx
7× H100 SXM 80 GB · 560 GB	Q4_K_M	454.4 GB	—	32K ctx	H100 SXM →	Q4_K_M · 454.4 GB · 32K ctx
4× H200 SXM 141 GB · 564 GB	Q4_K_M	454.4 GB	—	32K ctx	H200 SXM →	Q4_K_M · 454.4 GB · 32K ctx
3× B200 192 GB · 576 GB	Q4_K_M	454.4 GB	—	32K ctx	B200 →	Q4_K_M · 454.4 GB · 32K ctx

Also: Mac Studio (M3 Ultra, 512GB) · NVIDIA DGX A100 640GB · NVIDIA DGX H100

≤ 640 GB tier →

By VRAM tier · typical GPUs and Macs at Q4_K_M

Consumer GPUs (up to 32 GB) and Macs up to 192 GB: too small for this model (needs ≥ 213.1 GB)8 GB tier →12 GB tier →16 GB tier →24 GB tier →Macs →

How good is Mistral Large 3 675B Instruct 2512?

The llmrun Score merges 5 public benchmark boards into a single figure, where 100 represents the average for open models.

Score vs VRAM in the ≤512 GB tier

Mistral Large 3 675B Instruct 2512
Other models
Best Score for its VRAM
x: VRAM (GB, log scale) · y: llmrun Score

#71 of 102 that fit ≤512 GB

Each dot is a model at the best quant that fits ≤512 GB. The line steps up to the best Score available at each amount of VRAM; a dot below it is beaten by a model that needs no more memory.

Full ≤512 GB ranking →

Benchmarks

LMArena Text
LMArena · Oct 2026Arena
1414.3
#41/188 open
Arena
LMArena WebDev
LMArena · Oct 2026Arena
1232.6
#53/55 open
Arena
SciCode
Epoch AI · Dec 2025Lab-reported
36.6%
#47/72 open
Lab-reported
ALE-Bench
Epoch AI · Dec 2025Lab-reported
264.7
#46/48 open
Lab-reported
CritPt
Epoch AI · Dec 2025Lab-reported
0.0%
#72/84 open
Lab-reported
DTBench
Epoch AI · Dec 2025Lab-reported
65.1%
#39/70 open
Lab-reported

All 7 Mistral Large 3 675B Instruct 2512 benchmarks

In the ≤512 GB class

102 models fit ≤512 GB

Same Score, only models that fit ≤512 GB at their best quant: this model's peers by memory need.

#1
DeepSeek V4.1 Flash 1017DeepSeek
134.4
458.4 GB
#2
MiMo V2.6 Pro RLXiaomi
133.3
499.3 GB
#3
MiMo V2.6 Flash RLXiaomi
131.9
186.5 GB
#4
GLM 5.3Z.ai
130.1
453.8 GB
#5
DeepSeek V4.1 FlashDeepSeek
129.4
458.4 GB
#6
Step 5 PreviewStepFun
128.0
407.5 GB

Full ≤512 GB ranking →

Alternatives to Mistral Large 3 675B Instruct 2512

Mistral family · 35 models →

Official models with an llmrun Score, each answering one question; sizes are each model's default quant (usually Q4_K_M) with a 16K-token context.

Best alternative at this size
DeepSeek V4.1 Flash 1017
DeepSeek
- 44.8 higher Score
- 4.1 GB more VRAM
Score 134.4 · 458.4 GB
Compare with Mistral Large 3 675B Instruct 2512 Open DeepSeek V4.1 Flash 1017
Fits in less VRAM
GLM 5.3 Flash
Z.ai
- 34.5 higher Score
- 249.2 GB less VRAM
Score 124.1 · 205.2 GB
Compare with Mistral Large 3 675B Instruct 2512 Open GLM 5.3 Flash
Other vendor
GLM 5.3
Z.ai
- 40.5 higher Score
- 0.6 GB less VRAM
Score 130.1 · 453.8 GB
Compare with Mistral Large 3 675B Instruct 2512 Open GLM 5.3

Similar models

Model	Score	VRAM	tok/s	Why it is similar
DeepSeek V4.1 FlashDeepSeek	129.4	458.4 GB	—needs > 24 GB	similar score
Step 5 PreviewStepFun	128.0	407.5 GB	—needs > 24 GB	similar score
GLM 5.2Z.ai	122.9	453.8 GB	—needs > 24 GB	similar score
Beam 1.0 501B A23Breflection	119.5	304 GB	—needs > 24 GB	similar score
GLM 5.1Z.ai	117.8	454.1 GB	—needs > 24 GB	similar score
Qwen3.5 397B A17BAlibaba	112.8	242.6 GB	—needs > 24 GB	similar score
Qwen3.5 4BAlibaba	98.5	3.4 GB	~201	similar score
Llama 4 Maverick 17B 128E InstructMeta	87.1	272.7 GB	—needs > 24 GB	similar score
Llama 3.1 405B InstructMeta	83.5	275.6 GB	—needs > 24 GB	similar score
GLM 5Z.ai	111.8	454.1 GB	—needs > 24 GB	same size

Specifications

Publisher: Mistral AI
Family: Mistral
Parameters: 675B
Active parameters: 41B per token
Release Date: 2025-11-28
License: Apache 2.0

Resources

Frequently Asked Questions

How much VRAM does Mistral Large 3 675B Instruct 2512 need?

Mistral Large 3 675B Instruct 2512 requires 445.5 GB of VRAM at Q4_K_M, or 1485 GB at BF16.

VRAM = Weights + KV Cache + Overhead

Weights = 675B × 4.8 bits ÷ 8 = 405 GB

KV Cache + Overhead ≈ 40.5 GB (at 2K context + ~0.3 GB framework)

Fit ratings and hardware model lists check this model with room for a 16K-token context, which needs a little more memory.

VRAM usage by quantization

Q4_K_M

445.5 GB

Learn more about VRAM estimation →

Can NVIDIA GeForce RTX 5090 run Mistral Large 3 675B Instruct 2512?

No — Mistral Large 3 675B Instruct 2512 requires at least 204.2 GB at IQ2_XXS, which exceeds the NVIDIA GeForce RTX 5090's 32 GB of VRAM.

What's the best quantization for Mistral Large 3 675B Instruct 2512?

For Mistral Large 3 675B Instruct 2512, Q4_K_M (445.5 GB) offers the best balance of quality and VRAM usage. Q5_K_S (510.5 GB) provides better quality if you have the VRAM. The smallest option is IQ2_XXS at 204.2 GB.

VRAM requirement by quantization

IQ2_XXS

204.2 GB~53%

IQ3_XS

306.3 GB~73%

Q3_K_L

380.5 GB~86%

Q4_K_M ★

445.5 GB~89%

Q5_K_S

510.5 GB~92%

BF16

1485.0 GB~100%

★ Recommended — best balance of quality and VRAM usage.

Learn more about quantization →

Can I run Mistral Large 3 675B Instruct 2512 on a Mac?

Yes — Mac Studio (M3 Ultra, 256GB) and 1 other Mac can run Mistral Large 3 675B Instruct 2512, the smallest at IQ2_S. Apple Silicon uses unified memory, so the model shares RAM with the system. At Q4_K_M you need at least 445.5 GB of usable unified memory (RAM minus macOS overhead), from Mac Studio (M3 Ultra, 512GB).

Can I run Mistral Large 3 675B Instruct 2512 locally?

Yes — Mistral Large 3 675B Instruct 2512 can run locally on consumer hardware. Popular tools include Ollama, LM Studio, and llama.cpp.

What's the download size of Mistral Large 3 675B Instruct 2512?

At Q4_K_M, the download is about 405.00 GB. The full-precision BF16 version is 1350.00 GB. The smallest option (IQ2_XXS) is 185.63 GB.

Which GPUs can run Mistral Large 3 675B Instruct 2512?

No single consumer GPU has enough VRAM to run Mistral Large 3 675B Instruct 2512 at Q4_K_M (445.5 GB). Multi-GPU or professional hardware is required.

Which devices can run Mistral Large 3 675B Instruct 2512?

1 device with unified memory can run Mistral Large 3 675B Instruct 2512 at Q4_K_M (445.5 GB), including Mac Studio (M3 Ultra, 512GB). Apple Silicon Macs use unified memory shared between CPU and GPU, making them well-suited for local LLM inference.