This website uses cookies for authentication, support chat, and analytics. Only authentication cookies are turned on by default.
Animate this data
Bar chart race of Mistral AI, Google DeepMind, OpenAI and 7 more, by GPQA Diamond.
Create it free
Takes a free account
Mistral AI
Mistral AI has risen from 30.6% in 2023 to 88.1% in 2026.
Adversarial Nli
55 %
Aider Polyglot
11 %
Ale bench
1,213
Apex Agents
64 %
Arc Agi
6 %
BALROG
18 %
Bbh
56 %
Bool Q
87 %
Chess Puzzles
33 %
Critpt
11 %
Dtbench
88 %
Enigma Eval
1 %
Forecastbench
62 %
FrontierCode
39 %
FrontierMath
0 %
FrontierMath Tier 4 V2
27 %
FrontierMath Tiers 1 3 V2
59 %
Furniture Assembly
21 %
GPQA Diamond
88 %
Gsm8k
84 %
Hella Swag
87 %
Hle
5 %
Lech Mazur Writing
8 %
Live bench
48 %
Lmca
36 %
Math Level 5
82 %
Mmlu
80 %
Mystery Game Puzzles
17 %
Open Book Qa
86 %
OTIS Mock AIME 2024 2025
98 %
Piqa
84 %
Proofbench
10 %
SciCode
54 %
SimpleBench
23 %
SimpleQA Verified
20 %
Surface Evolver bench
28 %
Trivia Qa
82 %
WebDev Arena
1,265
WeirdML
44 %
WeirdML V3
4 %
Wino Grande
77 %
Training Compute
2,340,000,000,000.00T FLOP
Mistral AI benchmark results over time.