Mistral Small 3
By Mistral AI · France
Updated 2026-07-13
Overview
Mistral AI's 24B dense model that closes most of the gap with 70B-class models. Best quality-per-parameter we've measured at this size in 2025.
When to pick this model
- Self-hosting a near-frontier assistant on a single 24GB GPU
- Agentic workflows and tool calling where latency matters
- Long-context RAG with up to 128k tokens
- Commercial deployments needing Apache 2.0
- Replacing Llama 3 70B to cut VRAM and inference cost
VRAM requirements by quantization
| Quantization | VRAM required |
|---|---|
| Q4_K_M (recommended) | 14 GB |
| Q5_K_M | 17 GB |
| Q8_0 | 26 GB |
| FP16 (no quantization) | 48 GB |
VRAM figures include model weights plus a typical 8k KV cache and ~600 MB runtime overhead (Ollama / llama.cpp baseline). Add headroom for higher context lengths.
In practice, Mistral Small 3 needs a 16 GB card at Q4_K_M (14 GB). Stepping up to Q8_0 nearly doubles the footprint to 26 GB, and unquantized FP16 weights take 48 GB — plan your GPU around the Q4 or Q5 figure unless you specifically need the higher fidelity.
Without a GPU, Mistral Small 3 needs roughly 24 GB of system RAM to run on CPU via llama.cpp or Ollama — workable for background jobs, but far slower than GPU inference. Throughput estimates from our compatibility engine: around 4 tokens/sec on entry-level GPUs, on the order of 15 tokens/sec on a mid-range card, and up to 40 tokens/sec on high-end hardware — assuming the chosen quantization fully fits in VRAM.
What hardware do you need
The table below matches Mistral Small 3 to common GPU memory tiers, using the highest-fidelity quantization that fully fits each card class. Spilling layers to system RAM works but costs most of the speed, so size your card to the quantization you actually want to run.
| GPU memory | Example cards | Best fit for Mistral Small 3 |
|---|---|---|
| 8 GB | RTX 5070 Laptop, RTX 5060, RTX 5060 Ti 8GB | Does not fit — needs 14 GB at Q4_K_M |
| 12 GB | RTX 5070, RTX 5070 Ti Laptop, RTX 4080 Laptop | Does not fit — needs 14 GB at Q4_K_M |
| 16 GB | RTX 5080, RTX 4080 Super, Radeon RX 9070 XT | Q4_K_M (14 GB used) |
| 24 GB | RTX 4090, Radeon RX 7900 XTX, RTX 5090 Laptop | Q5_K_M (17 GB used) |
| 32 GB | RTX 5090 | Q8_0 (26 GB used) |
Which GPU should you buy to run Mistral Small 3?
To run Mistral Small 3 locally at Q4, you need ~14 GB of VRAM. The best value for this is a RTX 5070 Ti (16 GB VRAM).
As an Amazon Associate, BestLLMfor earns from qualifying purchases, at no extra cost to you. It does not influence our independent rankings.
Published benchmark scores
| Benchmark | Score |
|---|---|
| MMLU | 81 |
| GPQA | 42.2 |
| HumanEval | 84.8 |
Scores published by the model author or aggregated from public leaderboards. Re-measured monthly by our editorial team.
To put Mistral Small 3 in context: its MMLU score of 81 ranks #8 of the 34 catalog models with a published MMLU result (catalog median 73.4); its HumanEval score of 84.8 ranks #9 of the 26 catalog models with a published HumanEval result (catalog median 81.1). Scores are author-published and measured under different harnesses, so treat gaps of a few points as noise rather than a verdict.
Strengths
- Quality approaching Llama 3 70B at a third the size
- Low latency relative to peers
- 128k context window
- Strong tool use and agent behavior
- Apache 2.0 license
Limitations
- Needs ~16GB VRAM at Q4, more for higher precision
- Trails Qwen 2.5 Coder on dedicated coding tasks
- No native vision (see Small 3.1 for that)
Typical workloads
In our catalog grid, Mistral Small 3 is filed under Advanced Chat, Writing, Agents — the use cases where its size/quality trade-off makes the most sense. Its tags translate to concrete workloads: code generation and review (pair it with an editor integration like Continue.dev or Cline).
The 32k-token context window covers long chats and mid-sized documents, though very large retrieval workloads will need chunking. The Apache 2.0 license is permissive, so shipping it inside a commercial product raises no special legal questions.
Architecture & training
Architecture: Dense Transformer · 40 layers · GQA + sliding window
Training: Enriched multilingual corpus, strong focus on FR + scientific English.
The 2025 sweet spot for open-weight chat — frontier-adjacent quality at a tractable size.
Quick start
ollama run mistral-small:24bOr use the open-source MCP server to query this model from Claude Desktop, Cursor, or any MCP-compatible client.
Similar models worth comparing
Frequently asked questions
How much VRAM does Mistral Small 3 need?
At the recommended Q4_K_M quantization, Mistral Small 3 needs about 14 GB of VRAM. Q8_0 takes 26 GB, and unquantized FP16 weights take 48 GB.
Can Mistral Small 3 run without a GPU?
Yes — with roughly 24 GB of system RAM it runs CPU-only through llama.cpp or Ollama. Expect a fraction of GPU speed, which is fine for background or batch jobs but slow for interactive chat.
What context window does Mistral Small 3 support?
Mistral Small 3 supports a 32k-token context window (32,768 tokens).
Can I use Mistral Small 3 commercially?
Yes. Mistral Small 3 is released under Apache 2.0, a permissive open-source license that allows commercial use, modification and redistribution.
How fast is Mistral Small 3 on consumer hardware?
Our compatibility engine estimates on the order of 15 tokens/sec on a mid-range GPU and up to 40 tokens/sec on high-end cards, assuming the quantization fully fits in VRAM.
Which quantization of Mistral Small 3 should I download first?
Start with Q4_K_M (14 GB) — the standard size/quality sweet spot. Step up to Q5_K_M or Q8_0 only if you have VRAM headroom. On a 24 GB card you can run up to Q5_K_M.