dexiio

What hardware runs Mistral Large 3?

Mistral Large 3 is a 673.42B-parameter mixture-of-experts model with 39.95B active parameters (apache-2.0 license). Mistral's 675B MLA+MoE flagship. European frontier weight class — and a cloud referral for all but server rigs.

All figures assume an f16 KV cache, a 0.6 GB display reserve on the GPU, and 64 GB of DDR5 system RAM for the offload tiers. Tune these in the calculator.

Minimum VRAM by quant and context

QuantFile size8K16K32K64K128K
Q2_K *230.1 GB234 GB234.8 GB235.8 GB238 GB242.3 GB
Q3_K_M299.7 GB303.6 GB304.4 GB305.4 GB307.6 GB311.9 GB
Q4_K_M379.0 GB382.9 GB383.7 GB384.7 GB386.9 GB391.2 GB
Q5_K_M445.1 GB449 GB449.8 GB450.8 GB453 GB457.3 GB
Q6_K515.3 GB519.2 GB519.9 GB521 GB523.2 GB527.4 GB
Q8_0666.5 GB670.4 GB671.1 GB672.2 GB674.4 GB678.7 GB
IQ4_XS335.2 GB339.1 GB339.9 GB340.9 GB343.1 GB347.4 GB

Full-GPU figures: weights + f16 KV cache + overhead. * below our recommended floor of Q3_K_M.

GPU compatibility

GPU8K16K32K64K128K
NVIDIA GeForce RTX 5090
NVIDIA GeForce RTX 5080
NVIDIA GeForce RTX 5070 Ti
NVIDIA GeForce RTX 5070
NVIDIA GeForce RTX 5060 Ti 16GB
NVIDIA GeForce RTX 5060
NVIDIA GeForce RTX 4090
NVIDIA GeForce RTX 4080 SUPER
NVIDIA GeForce RTX 4070 Ti SUPER
NVIDIA GeForce RTX 4070
NVIDIA GeForce RTX 4060 Ti 16GB
NVIDIA GeForce RTX 4060
NVIDIA GeForce RTX 3090
NVIDIA GeForce RTX 3080 10GB
NVIDIA GeForce RTX 3070
NVIDIA GeForce RTX 3060 Ti
NVIDIA GeForce RTX 3060 12GB
NVIDIA GeForce RTX 2080 Ti
NVIDIA GeForce GTX 1080 Ti
AMD Radeon RX 9070 XT
AMD Radeon RX 7900 XTX
AMD Radeon RX 7900 XT
AMD Radeon RX 7800 XT
AMD Radeon RX 7600 XT
AMD Radeon RX 6800 XT
AMD Radeon RX 6700 XT
Intel Arc B580
Intel Arc A770 16GB
Apple M2 (16GB)
Apple M2 Pro (32GB)
Apple M2 Max (64GB)
Apple M2 Ultra (128GB)
Apple M3 (16GB)
Apple M3 Pro (36GB)
Apple M3 Max (64GB)
Apple M3 Ultra (96GB)
Apple M4 (16GB)
Apple M4 Pro (48GB)
Apple M4 Max (64GB)
CPU only / integrated graphics

Fits on GPUExpert offloadPartial offloadCPU only

Quant guidance

Our floor for Mistral Large 3 is Q3_K_M — below that, quality degrades faster than the VRAM savings are worth (large models tolerate low bpw better than small ones, which is why the floor is lower here). Prefer the highest quant that still lands "Fits on GPU" or “Expert offload” at your context length in the table above.

Frequently asked questions

How much VRAM does Mistral Large 3 need?

At Q3_K_M and 32K context, Mistral Large 3 needs about 305.4 GB of VRAM to run fully on GPU (weights + KV cache + overhead). As a mixture-of-experts model it can also run with far less VRAM via expert offload, keeping experts in system RAM.

What is the smallest GPU that can run Mistral Large 3?

No single consumer GPU in our set runs Mistral Large 3 well — it's a cloud-rental or multi-GPU model.

What quantization should I use for Mistral Large 3?

We recommend Q3_K_M or higher. Q4_K_M weighs 379.0 GB (4.83 bits/weight); going below Q3_K_M costs noticeable quality on a model this size.

How long a context can Mistral Large 3 handle?

Mistral Large 3 supports up to 288K tokens. KV cache grows linearly with context, though its MLA attention compresses the KV cache dramatically.

Can I run Mistral Large 3 without a GPU?

Not realistically — even at 8K context the weights don't fit in 64 GB of system RAM.

Related guides