Model

Mistral Small 3.1 (24B)

OPEN

Mistral AI · Mistral Small family · released Mar 17, 2025

First Mistral Small with native vision-language support; 128k context, Apache 2.0, fits a single RTX 4090 quantized.

ReasoningCodingVisionFunction callingTool useAgentic

1051.4

Elo · rank #204

Parameters

24B

Active params

24B (dense)

Context

128K tokens

Architecture

Dense transformer — 24B + Pixtral-style vision encoder, Tekken tokenizer

License

Apache 2.0

Languages

—

API price (in/out)

$0.1 / $0.3

Modalities

text · vision

Benchmark results

Bar shows position within the tracked field; marker = field best

AI2DVision93.7%#13

best: Molmo 72B · 96.3%

ChartQAVision86.2%#18

best: MiniMax-VL-01 · 91.7%

DocVQAVision94.1%#12

best: Qwen2-VL-72B · 96.5%

GPQA DiamondReasoning46.0%#203

best: GPT-5.6 · 94.6%

HumanEvalCoding88.4%#26

best: Claude Opus 4.5 · 99.4%

MATH-500Math69.3%#96

best: GPT-5 · 99.4%

MathVistaVision68.9%#30

best: Seed 2.1 Pro · 90.7%

MBPPCoding74.7%#26

best: Llama-3.3-Nemotron-Super-49B v1 (Reasoning On) · 91.3%

MMLU-ProKnowledge66.8%#101

best: Claude Fable 5 · 91.5%

MMLUKnowledge80.6%#84

best: OpenAI o3 · 92.9%

MMMU-ProVision49.3%#45

best: Claude Opus 4.7 · 85.5%

MMMUVision64.0%#68

best: Claude Fable 5 · 89.3%

SimpleQAKnowledge10.4%#36

best: GPT-4.5 · 62.5%

Run it locally

VRAM @ Q4

15 GB

VRAM @ FP16

55 GB

Fits on (Q4)

RTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4 · AWQ · MLX

Mistral Small family

Elo progression across releases

Mistral Small 3 (24B)Jan 2025978.8 Mistral Small 3.1 (24B)Mar 20251051.4 Mistral Small 3.2 (24B)Jun 20251075.6

Resources

HFmistralai/Mistral-Small-3.1-24B-Instruct-2503 GHmistralai/mistral-inference DOCSdocs.mistral.ai/models/mistral-small-3-1-25-03

API price $0.1/$0.3 · each benchmark row carries its own source badge (see methodology)