whichLlmmodel
Back to Dashboard
Text ModelOpen Source
Text

Codestral 22B VRAM Requirements

Developed by Mistral AI

Find out exactly how much VRAM you need to run Codestral 22B locally. Calculate the memory footprint of different GGUF quantization variants (like Q4_K_M or Q8_0), estimate your context length KV cache VRAM footprint, and determine if your hardware supports a full GPU VRAM offload or if you will need to rely on slow partial CPU offloading to avoid a CUDA Out of Memory (OOM) error.

Hugging Face Repository

Hardware Configuration

Adjust settings to check compatibility with your system in real time.

GB VRAM

Available: 14.50 GB

GB RAM

Available: 29.00 GB

8,192 tokens
CPU Offloading
Compatibility Verdict
Fits with CPU Offload

KV Cache & Overhead (3.0 GB) plus 11.5 GB weights fit in GPU VRAM (14.5 GB VRAM used). Remaining 1.8 GB weights are offloaded to System RAM (4.8 GB RAM required with OS).

Requires 16.25 GB total memory (weights: 13.30 GB, context overhead: 1.75 GB, activation overhead: 1.20 GB).

Memory Margin+27.25 GB
CPU Offload Memory Partitioning
GPU Dedicated VRAM
Model Weights11.55 GB
KV Cache1.75 GB
Activation & Runtime Overhead1.20 GB
Total VRAM Required14.50 GB
System RAM
Offloaded Model Weights1.75 GB
Total RAM Required1.75 GB

Codestral 22B Quantization Formats & VRAM Compatibility

Select a format to set it as active and calculate your system fit dynamically.

QuantWeightsKV CacheOverheadTotal VRAMStatusLinks
Base (Unquantized)41.44 GB1.75 GB3.46 GB
46.64 GB
Too Large
HF weights
Q4_K_M13.30 GB1.75 GB1.20 GB
16.25 GB
CPU Offload
GGUF
Q8_023.60 GB1.75 GB2.03 GB
27.38 GB
CPU Offload
GGUF

Codestral 22B KV Cache Memory Breakdown

How to Setup and Run Codestral 22B Locally

1

Method A: Ollama (Recommended)

Ollama is the easiest way to run models in the background. First, download it from ollama.com, then execute this terminal command:

ollama run <model-name>
2

Method B: LM Studio (GUI)

If you prefer a full graphical interface with chat UI and local server hosting:

  • Download and install LM Studio.
  • Search for Codestral 22B in the home page search tab.
  • Select a quantization level (like Q4_K_M) that fits your VRAM, click download, and load it to chat.

Model Specs

  • DeveloperMistral AI
  • Parameter Count22B
  • Base File Size41.4 GB
  • AvailabilityAPI & Local
  • Input Modalities
    Text
  • Official SiteVisit site

Standard Benchmark Scores

Coding (HumanEval)81.1%
Reasoning (GPQA Diamond)N/A

Commercial API Pricing

Input Tokens$0.30 / 1M
Output Tokens$0.90 / 1M

Model Release Alerts

Get notified when new models drop that run on your hardware. Zero spam.