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 DeepSeek-V3, gemini-4-argon_high, gpt-6-astra_max and 7 more, by GPQA Diamond.
Create it free
Takes a free account
DeepSeek-V3
Aider Polyglot
48 %
Bbh
88 %
Forecastbench
63 %
FrontierMath
2 %
GPQA Diamond
57 %
Hella Swag
89 %
Live bench
60 %
Math Level 5
65 %
Mmlu
87 %
OTIS Mock AIME 2024 2025
16 %
SciCode
36 %
SimpleBench
19 %
Trivia Qa
83 %
Wino Grande
85 %
Training Compute
3,300,000,000,000.00T FLOP
DeepSeek-V3 benchmark results.