Only with CPU offload

Mistral Small 4 119B-A6B at Q4_K_M needs 67.7 GB but only 7.0 GB is addressable, so about 91% of the layers would stream from system RAM at roughly 60 GB/s. Expect around 4.2 tokens per second — usable for batch work, painful for chat.

9% on GPU 8K context Q4_K_M · 66.9 GB Apache 2.0 Released 17 Mar 2026 Vision Not in the Ollama library

Instruct, reasoning, vision and code in one 119B MoE. A latent KV cache (320 wide) keeps context cheap; the weights still want 64 GB+.

What hardware do I need for Mistral Small 4 119B-A6B? →

The VRAM budget

weights 66.9 GB

Weights 66.9 GB KV cache @ 8K 0.18 GB Runtime overhead 0.6 GB Over budget 60.7 GB past 7.0 GB

Quantisation ladder

Quant	Weights	Total @ 8K	Max context	Tok/s	Quality	Fit
Q8_0	117.8 GB	118.5 GB	—	~2.3	−0.1% ppl	111.5 GB over
Q6_K	90.9 GB	91.7 GB	—	~3.0	−0.4% ppl	84.7 GB over
Q5_K_M	78.5 GB	79.3 GB	—	~3.5	−0.8% ppl	72.3 GB over
Q4_K_M	66.9 GB	67.7 GB	—	~4.2	−1.9% ppl	60.7 GB over
Q3_K_M	54.2 GB	54.9 GB	—	~5.2	−5.4% ppl	47.9 GB over
Q2_K	46.4 GB	47.2 GB	—	~6.2	−15% ppl	40.2 GB over

Quality is the published perplexity delta against f16 weights. Max context assumes an f16 KV cache; q8_0 roughly doubles it. This model uses multi-head latent attention, so its cache is a compressed latent rather than full K and V.

How to run it

llama.cpp LM Studio vLLM

terminal

$ llama-server \
    -hf mistralai/Mistral-Small-4-119B-2603:Q4_K_M \
    -c 8192 -ngl 3

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

01Download is 66.9 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