Mistral Small 3.1 24B
ToolsVisionPopularMistral AI · 24B · Dense
Multimodal Mistral with vision support
156.5K
1.4K
2025-03
128K tokens
Use cases
Your hardware
16 GB Reference Desktop
Quantization options
What is quantization?| Quant | Bits | VRAM | Disk | Quality | Status |
|---|---|---|---|---|---|
| Q2_K | 2 | 8.2 GB | 7.3 GB | low | Decent · ~25 t/s |
| Q3_K_M | 3 | 11.3 GB | 10.3 GB | moderate | Tight fit · ~18 t/s |
| Q4_K_M | 4 | 12.8 GB | 11.7 GB | good | Tight fit · ~16 t/s |
| Q5_K_M | 5 | 15.9 GB | 14.7 GB | good | Barely runs · ~13 t/s |
| Q6_K | 6 | 18.9 GB | 17.6 GB | excellent | Barely runs · ~5 t/s |
| Q8_0 | 8 | 25.1 GB | 23.5 GB | excellent | Barely runs · ~3 t/s |
| F16 | 16 | 49.7 GB | 46.9 GB | lossless | Too heavy · ~1 t/s |
About this model
Mistral Small 3.1 24B is a 24B dense transformer from Mistral AI in the Mistral family, designed primarily for chat, vision, code workflows. Supports structured tool calling; use Q4_K_M or higher when strict JSON schema adherence matters. Supports up to 128K context tokens; start at 4K–8K with Q8_0 KV cache on tight GPUs.
Hardware planning baselines start at 13.4 GB minimum memory and 22.4 GB recommended memory. At Q4_K_M, weights plus baseline runtime overhead use roughly 12.8 GB of RAM/VRAM and 11.7 GB of disk space. As a 24B dense checkpoint, every parameter participates in each token forward pass, meaning decoding speed scales linearly with GPU memory bandwidth and drops sharply if layers spill over PCIe into host DDR5 RAM.
Deploy Mistral Small 3.1 24B locally via Ollama (`mistral-small`), LM Studio, or Ollama with Flash Attention enabled. Under Apache License 2.0, review official weight cards before commercial redistribution.
Highlights
- Local Fit Baseline13.4 GB minimum and 22.4 GB recommended RAM/VRAM.
- Primary QuantQ4_K_M targets 12.8 GB memory and 11.7 GB disk.
- Context & Capabilities128K max context · Dense · Tool calling.
System requirements by device
Memory planning targets 13.4 GB minimum RAM/VRAM and 22.4 GB recommended memory. The baseline quant is Q4_K_M, estimated at 12.8 GB memory and 11.7 GB disk. Actual runtime usage shifts with context length, KV cache precision, and OS background overhead.
| Device | Notes |
|---|---|
| Mac (Apple silicon) | Use unified memory headroom (~75% physical RAM max) as the primary fit signal. Start with Q4_K_M and close heavy browser tabs first. |
| Windows / Linux GPU | Keep weights plus KV cache inside dedicated VRAM so Windows WDDM does not spill layers over PCIe into Shared GPU Memory. |
| Phones & Compact Edge Boards | Not practical on phones (13.4 GB min RAM exceeds Android LMK and iOS Jetsam limits). Pick a 2B–4B model on mobile. |
Start with a short session and leave memory for the operating system. A file fitting on disk does not guarantee that it will load in memory.
Estimated memory by quantization
These are rough estimates, not tested minimum requirements. Check the actual file and runtime documentation; context, offloading and operating-system overhead can change the memory needed.
| Quant Tier | Estimated memory | Disk Size | Quality |
|---|---|---|---|
| Q2_K | 8.2 GB | 7.3 GB | low |
| Q3_K_M | 11.3 GB | 10.3 GB | moderate |
| Q4_K_M | 12.8 GB | 11.7 GB | good |
| Q5_K_M | 15.9 GB | 14.7 GB | good |
| Q6_K | 18.9 GB | 17.6 GB | excellent |
| Q8_0 | 25.1 GB | 23.5 GB | excellent |
| F16 | 49.7 GB | 46.9 GB | lossless |
How to run Mistral Small 3.1 24B on your device
Start with a short prompt and the smallest quant that still gives the quality you need. A clean first run is more useful than loading a larger file with no memory headroom.
Run Mistral Small 3.1 24B on Mac
- Install Ollama or LM Studio on macOS Apple Silicon.
- Pull Q4_K_M (12.8 GB) first and confirm it stays below ~75% of your Mac's unified memory.
- Enable Flash Attention and cap initial context at 8K until memory pressure stays below 22.4 GB.
- Consult the Mac guide for Activity Monitor tuning and Metal offload tips.
Run Mistral Small 3.1 24B on Windows
- Install LM Studio for Windows for visual VRAM inspection, or Ollama when you want a localhost API endpoint.
- Load Q4_K_M (~12.8 GB) and verify Dedicated GPU Memory stays within VRAM without spilling into Shared GPU Memory.
- Reserve at least 13.4 GB memory minimum and target 22.4 GB for comfortable context growth.
- Enable Q8_0 KV cache or drop one quant tier if generation slows down over PCIe.
Mistral Small 3.1 24B vs alternatives
| Model | Params | Context | Min RAM | License |
|---|---|---|---|---|
| Mistral Small 3.1 24B | 24B | 128K | 13.4 GB | Apache 2.0 |
| Devstral Small 2 24B | 24B | 256K | 13.4 GB | Apache 2.0 |
| Ministral 3 14B | 14B | 256K | 7.8 GB | Apache 2.0 |
| Mistral Nemo 12B | 12B | 128K | 6.7 GB | Apache 2.0 |
| Ministral 8B | 8B | 32K | 4.5 GB | MRL |
| Mistral 7B v0.3 | 7B | 32K | 3.9 GB | Apache 2.0 |
Use this table to choose the smallest nearby model that still matches the job. A related model with lower RAM can be the better local choice when the current page fits only at the edge of available memory.
Frequently asked questions
- What is Mistral Small 3.1 24B?
- Mistral Small 3.1 24B is a 24B dense transformer from Mistral AI, supporting up to 128K tokens with 13.4 GB minimum RAM and 22.4 GB recommended RAM.
- How much VRAM or RAM does Mistral Small 3.1 24B require?
- At Q4_K_M, Mistral Small 3.1 24B occupies roughly 12.8 GB of active memory and 11.7 GB on disk. Add 2–6 GB of spare headroom for KV cache buffers, runtime scratch space, and OS apps (22.4 GB recommended).
- What is the best way to run Mistral Small 3.1 24B locally?
- Use `ollama run mistral-small` for a quick CLI/API setup, or load Q4_K_M in LM Studio or llama.cpp with Flash Attention enabled when you want explicit KV cache and layer control.
- Can Mistral Small 3.1 24B run on Android or iPhone?
- Mistral Small 3.1 24B requires 13.4 GB minimum memory, placing it outside practical 8 GB smartphone limits. Run it on a 16 GB+ Mac or desktop GPU instead, and pick a 2B–4B checkpoint for phones.
- How does Mistral Small 3.1 24B compare to Devstral Small 2 24B?
- Devstral Small 2 24B (24B, 13.4 GB min RAM) sits closest in size. Compare active parameters, context ceiling (128K vs 256K), and license terms (Apache 2.0) before switching.