Model

Pixtral 12B

OPEN

Mistral AI · Pixtral family · released Sep 11, 2024

Mistral's first multimodal model; 12B decoder + 400M vision encoder, 128k context, Apache 2.0; now retired from API.

ReasoningCodingVisionFunction callingTool useAgentic

617.0

Elo · rank #281

Parameters

12.4B

Active params

12.4B (dense)

Context

128K tokens

Architecture

Dense decoder (12B) + vision encoder (0.4B)

License

Apache 2.0

Languages

—

API price (in/out)

No hosted API

Modalities

text · vision

Benchmark results

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

ChartQAVision81.8%#28

best: MiniMax-VL-01 · 91.7%

DocVQAVision90.7%#31

best: Qwen2-VL-72B · 96.5%

HumanEvalCoding72.0%#77

best: Claude Opus 4.5 · 99.4%

IFEvalReasoning61.3%#120

best: Gemma 4 26B A4B · 98.5%

MATH-500Math48.1%#138

best: GPT-5 · 99.4%

MathVistaVision58.0%#46

best: Seed 2.1 Pro · 90.7%

MMLUKnowledge69.2%#151

best: OpenAI o3 · 92.9%

MMMUVision52.5%#91

best: Claude Fable 5 · 89.3%

MT-BenchHuman preference7.68#52

best: Hunyuan-Large (A52B) · 9.4

VQAv2Vision78.6%#10

best: Reka Edge (2026) · 88.4%

Run it locally

VRAM @ Q4

7.5 GB

VRAM @ FP16

25 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 · MLX

Pixtral family

Elo progression across releases

Pixtral 12BSep 2024617.0 Pixtral Large (124B)Nov 20241190.3

Resources

HFmistralai/Pixtral-12B-2409 DOCSdocs.mistral.ai/models/overview

API price weights · each benchmark row carries its own source badge (see methodology)