Skip to content

Ministral 3 8B on RTX 5090: VRAM, speed and fit

Ministral 3 8B at 8-bit runs on the RTX 5090 with 152K of context at about 123 tokens a second.

Text generation · Chat · Ollama / LM Studio

Runs comfortably · on RTX 5090 · 8-bit (Q8_0) · Ollama / LM Studio · Chat

GPU memory
29.8 GB
of 32.0 GB
Context
152K
auto
Speed
123 tok/s
39 when full
First reply
0.2 s
2K prompt

Where the memory goes

Model
8.8 GB
Context
20.2 GB
Overhead
845 MB
OS
0.0 GB
Free
2.2 GB
Memory
29.8 GB of 32.0 GB
Context
152K tokens · auto
Speed
123 tok/s · 39 when full
First reply
0.2 s for a 2K prompt
Running cost
$9.30/mo at home · $58/mo on a cloud 2× RTX 2000 Ada

Validation: Validated

What Ministral 3 8B on RTX 5090 was checked against
CheckTargetResultStatus
Fits in GPU memory32.0 GB usable29.8 GB neededPass
Headroom5% free2.2 GB free (7%)Pass
Context for chat16K or more152KPass
Writing speed15 tok/s or more123 tok/sPass
First replyunder 3 s, 2K prompt0.2 sPass

Targets: a reply you can read along with, a first reply the use case tolerates, and 5% of memory left free. Calculated from published specs, not benchmarked.

The setup

Model
Ministral 3 8B
Parameters
8.9B
Precision
8-bit (Q8_0)
Context
152K, auto
Chats at once
1
Use
Chat
Runs in
Ollama / LM Studio
Hardware
RTX 5090
GPU can use
32.0 GB of 32.0 GB

Running cost

At home
$9.30/mo
425 W at the wall
Renting instead
$58/mo
2× RTX 2000 Ada at $0.48/h

4 h a day, power at $0.18 per kWh, cloud at RunPod on-demand prices as of Nov 22, 2026.

How to run it

Ollama
ollama run ministral-3:8b
llama.cpp
llama-server -hf unsloth/Ministral-3-8B-Instruct-2512-GGUF
vLLM
vllm serve mistralai/Ministral-3-8B-Instruct-2512
MLX (Mac)
mlx_lm.chat --model mlx-community/Ministral-3-8B-Instruct-2512-4bit

Ministral 3 8B on similar machines

All machines: Ministral 3 8B VRAM requirements. Everything that runs here: what can the RTX 5090 run?

Other models that run on RTX 5090

Questions

Can Ministral 3 8B run on RTX 5090?

Ministral 3 8B at 8-bit runs on the RTX 5090 with 152K of context at about 123 tokens a second.

Report VR-0NQFT2M, calculated Nov 22, 2026 from published specs, not benchmarked.