Mistral AI

Mistral Medium 3

Type
language model
Context
131K tokens
Max output
128K
Released
7 May 2025
API string
mistral-medium-2505

Prices by provider

Provider Input, $ per 1M Output, $ per 1M In our data since
Azure Cognitive Services $0.4 $2 31 Jul 2026
Merge Gateway $0.4 $2 31 Jul 2026
Microsoft Foundry $0.4 $2 31 Jul 2026
Mistral AI $0.4 $2 31 Jul 2026
IBM watsonx.ai $3 $10 31 Jul 2026

Only the base tier and only per-token prices are shown. Batch, discounted and cached rates, as well as prices per image or per second of video, do not go into this table: they cannot stand in the same column as a price per million tokens.

The date in the last column is the day this price first entered our collection. The price may well be older: before that day we simply were not recording it. It has not changed since — otherwise a new row with a new date would stand in its place.

Measurement results

Task set Result Run conditions Measured by
Arena Score, programming 1,386.17 1,378–1,394
Arena Score in English 1,384.32 1,379–1,390
Arena Score, multi-turn dialogue 1,383.26 1,375–1,391
Arena Score in French 1,370.05 1,342–1,398
Arena Score, overall 1,369.43 1,365–1,374
Arena Score in Spanish 1,365.42 1,338–1,393
Arena Score, hard prompts 1,364.71 1,359–1,371
Arena Score, long queries 1,358.37 1,350–1,366
Arena Score in Russian 1,357.45 1,346–1,369
Arena Score, mathematics 1,351.89 1,340–1,364
Arena Score, creative writing 1,343.78 1,334–1,353
Arena Score, expert questions 1,343.67 1,330–1,357
Arena Score, instruction following 1,338.98 1,332–1,346
Arena Score, working with images 1,157.54 1,149–1,166
Arena Score, understanding diagrams 1,154.86 1,139–1,171
Arena Score, text recognition in images 1,153.86 1,144–1,164
MATH, difficulty level five 81.63 % · with a tuned harness Epoch evaluations
Creative writing (Lech Mazur’s evaluation) 77.3 % lechmazur/writing Github repository
GPQA Diamond — graduate-level questions 46.04 % · with a tuned harness Epoch evaluations
Mock AIME 2024–2025 — olympiad problems 32.15 % · with a tuned harness Epoch evaluations
Humanity’s Last Exam — expert-level questions 0 %