Model

Ministral 8B

OPEN

Mistral AI · Ministral family · released Oct 16, 2024

Open-weights 8B edge model with interleaved sliding-window attention (research license); instruct weights on HF.

ReasoningCodingVisionFunction callingTool useAgentic

553.7

Elo · rank #288

Parameters

8.02B

Active params

8.02B (dense)

Context

128K tokens

Architecture

Dense Transformer w/ interleaved sliding-window attention, 36 layers

License

Mistral Research License (non-commercial; commercial license required for production)

Languages

—

API price (in/out)

$0.1 / $0.1

Modalities

text

Benchmark results

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

AGIEvalReasoning48.3%#20

best: OLMo 3-Think 32B · 88.2%

ARC-ChallengeReasoning71.9%#44

best: Llama 3.1 405B · 96.9%

Arena-HardHuman preference70.9%#34

best: Qwen3-235B-A22B (Non-Thinking) · 96.1%

GSM8KMath64.5%#108

best: Llama 3.1 405B · 96.8%

HumanEvalCoding76.8%#61

best: Claude Opus 4.5 · 99.4%

MATH-500Math54.5%#120

best: GPT-5 · 99.4%

MBPPCoding70.0%#39

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

MMLUKnowledge65.0%#179

best: OpenAI o3 · 92.9%

MT-BenchHuman preference8.3#37

best: Hunyuan-Large (A52B) · 9.4

TriviaQAKnowledge65.5%#27

best: Sarvam-1 (2B) · 90.6%

WildBenchHuman preference41.3%#8

best: Mistral Small 3.2 (24B) · 65.3%

WinoGrandeReasoning75.3%#63

best: PaLM 2 · 90.9%

Run it locally

VRAM @ Q4

5 GB

VRAM @ FP16

16 GB

Fits on (Q4)

RTX 3060 12GBRTX 4070 Ti 16GBRTX 3090 24GBRTX 4090 24GBRTX 5090 32GBM4 Pro 48GBM3 Max 128GBM3 Ultra 512GBA100 80GBH100 80GBH200 141GBB200 192GB

Throughput data unavailable.

Quantizations

GGUF Q4 · AWQ · MLX

Ministral family

Elo progression across releases

Ministral 3BOct 2024423.7 Ministral 8BOct 2024553.7

Resources

HFmistralai/Ministral-8B-Instruct-2410 DOCSdocs.mistral.ai/models/overview

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