Model
Fugu
Multi-agent model for routing expert agents across complex analytical tasks
1 providersreleased closed weights
Compare providers (sorted by listed input price)
| Provider | Input /M | Output /M | Cache read /M | Cache write /M | Context | Capabilities | Status | Updated |
|---|---|---|---|---|---|---|---|---|
| Sakana AI fugu | — | — | — | — | 1M | reasoningtoolsstructuredvision | Jun 15, 2026 |
Benchmarks
- SWE Bench Pro ↗report ↗59
- Terminal Bench 2.1 ↗report ↗80.2
- LiveCodeBench ↗report ↗92.9
- LiveCodeBench Pro ↗report ↗87.8
- Humanity’s Last Exam ↗report ↗47.2
- CharXiv Reasoningreport ↗85.1
- GPQA Diamond ↗report ↗95.5
- SciCodereport ↗60.1
- τ3 Bankingreport ↗21.7
- Long Context Reasoningreport ↗74.7
- MRCRv2report ↗86.6
- CTI-REALMreport ↗67.5
Prices are per million tokens (USD). “—” means the provider does not publicly list a price for this model. Data from models.dev; verify with the provider before purchasing.