Devstral 2

Available

Mistral · 2025-12-09 · 32,000 tokens

An AI model from Mistral, suited to a broad range of AI workloads.

Supported modalities:textcode

Quick Overview

Text Generation2/10

Code Generation3/10

Reasoning4/10

Multimodal2/10

Scores from leading benchmark suites.

artificial analysis intelligence19.2

artificial analysis coding31.3

artificial analysis math36.7

Latency and throughput performance.

—P50 Latency

45.318tokens/sec