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 Moonshot, Google DeepMind, OpenAI and 7 more, by GPQA Diamond.
Create it free
Takes a free account
Moonshot
Moonshot has risen from 84.2% in 2025 to 93.1% in 2026.
Aider Polyglot
59 %
Ale bench
1,933
Apex Agents
62 %
Arc Agi
95 %
Arc Agi 2
60 %
Blueprint bench 2
30 %
Chess Puzzles
37 %
Cl bench
19 %
Cl bench Life
13 %
Critpt
25 %
Deepswe
69 %
Dtbench
91 %
EBR bench
2 %
Enigma Eval
3 %
Forecastbench
65 %
FrontierCode
44 %
FrontierMath
39 %
FrontierMath Tier 4
15 %
FrontierMath Tier 4 V2
39 %
FrontierMath Tiers 1 3 V2
72 %
FrontierSWE
26 %
Furniture Assembly
34 %
GBAEval
48 %
GPQA Diamond
93 %
Hle
24 %
Lech Mazur Writing
9 %
Lmca
53 %
Mystery Game Puzzles
26 %
Os World
63 %
OTIS Mock AIME 2024 2025
97 %
PostTrainBench
32 %
Proofbench
91 %
SciCode
61 %
SimpleBench
61 %
SimpleQA Verified
51 %
Surface Evolver bench
95 %
SWE bench Verified
77 %
TerminalBench
43 %
Vending bench 2
6,205
WebDev Arena
1,674
WeirdML
83 %
WeirdML V3
8 %
Training Compute
20,001,000,000,000.00T FLOP
Moonshot benchmark results over time.