Devstral 2

Mistral·Released 2025-12-09

Open Source

Compare

Quality Index

22.0

228th of 507

Top 45%

Coding Index

23.7

168th of 417

Top 41%

Math Index

36.7

170th of 269

Top 63%

Speed

65 tok/s

Top 66%

TTFT

0.60s

Market Position

Devstral 2Market Average

Performance

65

tokens/sec

Faster than 34% of models

0.60

seconds

Faster than 75% of models

0.60

seconds

Faster than 83% of models

Market Median

86 tok/s

25% slower

Median TTFT

1.07s

44% faster

Speed Comparison

Qwen3 14B (Reasoning)

65 tok/s+0%

Apertus 70B Instruct

64 tok/s-0%

Gemma 3n E2B Instruct

64 tok/s-0%

Benchmarks

MMLU-Pro

76.2%

GPQA Diamond

59.4%

HLE

3.6%

LiveCodeBench

44.8%

SciCode

33.1%

TerminalBench Hard

18.9%

MATH-500Not evaluated

AIMENot evaluated

AIME 2025

36.7%

IFBench

38.1%

Long Context Recall

30.0%

Tau2

24.9%

Market AverageTop Score

Open Source

Quick Compare

Similar Models

GLM-4.7-Flash (Non-reasoning)

Z AI

Q: 22.1$0.15/1M

Faster: 38%Coding: -12.7

Prime Intellect: INTELLECT-3

Prime Intellect

Q: 22.2$0.43/1M131K ctx

Coding: -4.6

GPT-5 (ChatGPT)

OpenAI

Q: 21.8$3.44/1M

Faster: 160%

Solar Open 100B (Reasoning)

Upstage

Q: 21.7N/A/1M

Coding: -13.2

DeepSeek V3 0324

DeepSeek

Q: 22.3$1.21/1M

Gemini 2.5 Flash-Lite Preview (Sep '25) (Reasoning)

Google

Q: 21.6$0.17/1M

Faster: 437%Coding: -5.6

Compare all 7 models

Quality Index

22.0

228th of 507

Top 45%

Coding Index

23.7

168th of 417

Top 41%

Math Index

36.7

170th of 269

Top 63%

Speed

65 tok/s

Top 66%

TTFT

0.60s

Market Position

Devstral 2Market Average

Devstral 2

Related Models

Market Position

Performance

Benchmarks

Open Source

Quick Compare

Similar Models

Market Position