Sakana Fugu-Ultra — Real Scores Across 10 Benchmarks (2026)

Sakana AI

Sakana Fugu-Ultra

10 benchmarks

GPQA Diamond

#1 of 124

95.5%

MRCR v2

#2 of 18

93.6%

LiveCodeBench

#3 of 56

93.2%

LiveCodeBench Pro

#1 of 2

90.8%

CharXiv

#5 of 32

86.6%

Terminal-Bench 2.1

#14 of 34

82.1%

SWE-bench Pro

#4 of 49

73.7%

Long Context Reasoning

#1 of 1

73.3%

SciCode

#2 of 3

58.7%

Humanity's Last Exam

#21 of 75

50.0%