Sakana AI
Fugu
Multi-agent model for routing expert agents across complex analytical tasks
sakana/fugu
Context
1M
Output
—
Released
2026-06-15
Knowledge cutoff
—
Types
InputTextImage
OutputText
Reasoning
Supported
Tool use
Supported
Structured output
Supported
Attachments
Supported
Open weights
—
Provider pricing
No providers currently list this model.
Benchmarks
SWE Bench Pro
59
Terminal Bench 2.1
80.2
LiveCodeBench
92.9
LiveCodeBench Pro
87.8
Humanity’s Last Exam
47.2
CharXiv Reasoning
85.1
GPQA Diamond
95.5
SciCode
60.1
τ3 Banking
21.7
Long Context Reasoning
74.7
MRCRv2
86.6
CTI-REALM
67.5