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 Anthropic, Google DeepMind, OpenAI and 7 more, by GPQA Diamond.
Create it free
Takes a free account
Anthropic
Anthropic has risen from 34.7% in 2023 to 95.6% in 2026.
Aider Polyglot
72 %
Ale bench
2,165
Algotune
2 %
Apex Agents
84 %
Arc Agi
99 %
Arc Agi 2
93 %
BALROG
63 %
Blueprint bench 2
51 %
Cad Eval
54 %
Chess Puzzles
67 %
Cl bench
21 %
Cl bench Life
17 %
Critpt
32 %
CursorBench
58 %
DeepResearchBench
55 %
Deepswe
74 %
Dtbench
99 %
EBR bench
71 %
Enigma Eval
39 %
Forecastbench
67 %
FrontierCode
55 %
FrontierMath
47 %
FrontierMath Erdos
3 %
FrontierMath Tier 4
31 %
FrontierMath Tier 4 V2
95 %
FrontierMath Tiers 1 3 V2
91 %
FrontierSWE
62 %
Furniture Assembly
83 %
GBAEval
80 %
GDPval
46 %
GPQA Diamond
96 %
Gsm8k
87 %
Hle
47 %
Lech Mazur Writing
8 %
Live bench
76 %
Lmca
68 %
Math Level 5
98 %
Mindcube
45 %
MirrorCode
79 %
Mmlu
87 %
Mystery Game Puzzles
71 %
Os World
72 %
OTIS Mock AIME 2024 2025
100 %
PostTrainBench
42 %
Proofbench
100 %
Rli
21 %
SciCode
67 %
Science Qa
72 %
SimpleBench
82 %
SimpleQA Verified
72 %
Spatialviz bench
34 %
Spireclimb
3 %
Spireclimb Sts2 A10
7 %
Surface Evolver bench
95 %
SWE bench Verified
83 %
TerminalBench
80 %
Trivia Qa
88 %
Vending bench 2
11,182
WebDev Arena
1,820
WeirdML
93 %
WeirdML V3
31 %
Wino Grande
89 %
Training Compute
33,500,000,000,000.00T FLOP
Anthropic benchmark results over time.