Only with CPU offload

Ministral 3 14B at Q4_K_M needs 9.7 GB but only 7.0 GB is addressable, so about 34% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 11 tokens per second — usable for batch work, painful for chat.

66% on GPU 8K context Q4_K_M · 7.8 GB Apache 2.0 Released Dec 2025 Vision

The largest Ministral. A 12 GB card runs it at Q4 with a few gigabytes to spare.

Hardware requirements What hardware do I need to run Ministral 3 14B? →

At Q4_K_M with an 8K context: a 16 GB GPU (RTX 5080) or a 24 GB Mac (M4 Pro). Minimum, recommended, high-quality and long-context tiers, with all 114 devices checked at every quantisation.

Fits instead: Ministral 3 8B (6.7 GB) · Ministral 3 3B (3.6 GB)

The VRAM budget

weights 7.8 GB
Weights 7.8 GB KV cache @ 8K 1.25 GB Runtime overhead 0.6 GB Over budget 2.7 GB past 7.0 GB

Quantisation ladder

QuantWeightsTotal @ 8KMax contextTok/sQualityFit
F16 25.9 GB 27.7 GB — ~1.7 Reference 20.7 GB over
Q8_0 13.8 GB 15.6 GB — ~4.0 −0.1% ppl 8.6 GB over
Q6_K 10.6 GB 12.5 GB — ~6.1 −0.4% ppl 5.5 GB over
Q5_K_M 9.2 GB 11.0 GB — ~8.0 −0.8% ppl 4.0 GB over
Q4_K_M 7.8 GB 9.7 GB — ~11 −1.9% ppl 2.7 GB over
Q3_K_M 6.3 GB 8.2 GB — ~22 −5.4% ppl 1.2 GB over
Q2_K 5.4 GB 7.3 GB 6K ~47 −15% ppl 0.3 GB over

Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it.

How to run it

terminal
$ llama-server \
    -hf mistralai/Ministral-3-14B-Instruct-2512:Q4_K_M \
    -c 8192 -ngl 26

The engine underneath most of the others. Every knob is exposed. More on llama.cpp.

01Download is 7.8 GB. Keep it on an SSD — a first load off a spinning disk takes minutes.
02Close anything else holding VRAM. A browser with hardware acceleration can sit on 1–2 GB.
03If it falls back to CPU silently, drop the context first, then step down a quantisation.
See all models for this rig Compare with another model

Related models on a GeForce RTX 3070 Ti

ModelTotalTok/s est.
Ministral 3 8B
8.92B · same family
Q4_K_M · 6.7 GB 73 Tight fit
Ministral 3 3B
3.85B · same family
Q4_K_M · 3.6 GB 170 Runs great
Phi-4 14B
14.7B · similar size
Q4_K_M · 10.4 GB ~9.3 CPU offload
Qwen3 14B
14.8B · similar size
Q4_K_M · 10.2 GB ~9.9 CPU offload
DeepSeek-R1-Distill-Qwen 14B
14.8B · similar size
Q4_K_M · 10.4 GB ~9.3 CPU offload
Qwen2.5-Coder 14B
14.8B · similar size
Q4_K_M · 10.4 GB ~9.3 CPU offload
Gemma 3 12B
12.2B · similar size
Q4_K_M · 8.3 GB ~20 CPU offload

Every model against the GeForce RTX 3070 Ti →

Ministral 3 14B on other hardware

DeviceTotalTok/s est.
GeForce RTX 4060 Laptop
7.0 GB usable · 256 GB/s
9.7 GB ~9.4 CPU offload
GeForce RTX 3060 12 GB
10.6 GB usable · 360 GB/s
9.7 GB 28 Tight fit
GeForce RTX 5080
14.4 GB usable · 960 GB/s
9.7 GB 74 Runs great
M4 · 16 GB
10.7 GB usable · 120 GB/s
9.7 GB 8.6 Tight fit
GeForce RTX 4090
22.4 GB usable · 1008 GB/s
9.7 GB 78 Runs great
Radeon RX 7900 XTX
22.4 GB usable · 960 GB/s
9.7 GB 74 Runs great
M4 Pro · 24 GB
18.0 GB usable · 273 GB/s
9.7 GB 20 Runs great
M3 Max · 64 GB
48.0 GB usable · 400 GB/s
9.7 GB 29 Runs great

All 114 devices, by tier →