Model

Mistral Large 3

OPEN

Mistral AI · Mistral Large family · released Dec 2, 2025

Mistral's Dec 2025 flagship, not previously in corpus (corpus's newest Mistral entries were Magistral/Small 3.2 from June 2025). Apache 2.0, trained on 3000 H200s per vendor.

ReasoningCodingVisionFunction callingTool useAgentic

1303.2

Elo · rank #155

Parameters

675B

Active params

41B (MoE)

Context

256K tokens

Architecture

Sparse Mixture-of-Experts, 675B total / 41B active parameters

License

Apache 2.0

Languages

—

API price (in/out)

$0.5 / $1.5

Modalities

text · vision

Benchmark results

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

AGIEvalReasoning74.0%#3

best: OLMo 3-Think 32B · 88.2%

AIMEMath40.0%#126

best: GPT-5.2 · 100.0%

Arena EloHuman preference1418#17

best: Claude Fable 5 · 1505

GPQA DiamondReasoning67.2%#132

best: GPT-5.6 · 94.6%

Humanity's Last ExamReasoning4.0%#131

best: Claude Sonnet 5 · 57.4%

HumanEvalCoding90.2%#16

best: Claude Opus 4.5 · 99.4%

IFBenchReasoning36.0%#60

best: MiniMax M3 · 83.0%

LiveCodeBenchCoding55.3%#111

best: DeepSeek-V4-Pro (Think Max) · 93.5%

MATH-500Math93.6%#40

best: GPT-5 · 99.4%

MMLU-ProKnowledge73.1%#84

best: Claude Fable 5 · 91.5%

MMMU-ProVision56.0%#40

best: Claude Opus 4.7 · 85.5%

MMMUVision66.2%#64

best: Claude Fable 5 · 89.3%

SWE-bench VerifiedCoding41.4%#102

best: Claude Fable 5 · 95.0%

Terminal-Bench 2.0Coding12.0%#60

best: GPT-5.6 · 88.8%

Run it locally

VRAM @ Q4

—

VRAM @ FP16

—

Fits on (Q4)

Multi-node cluster required

Throughput data unavailable.

Quantizations

—

Mistral Large family

Elo progression across releases

Mistral LargeFeb 2024713.0 Mistral Large 2 (24.07)Jul 20241068.2 Mistral Large 24.11Nov 2024658.0 Mistral Large 3Dec 20251303.2

Resources

HFmistralai/Mistral-Large-3-675B-Instruct-2512 DOCShttps://docs.mistral.ai/APIAvailable via Mistral AI API

API price $0.5/$1.5 · each benchmark row carries its own source badge (see methodology)