Sakana AI · Fugu
Fugu Max
Lower-cost Sakana AI Fugu orchestration tier (pool includes open models such as Nemotron) positioned against mid-price frontier models.
- AA Intelligence
- —
- Input / output per 1M
- $2.00/ $6.00
- Output speed
- —
- Context window
- —
Coding
- Terminal-Bench 2.x#4 of 10389.5%
Leader: Claude Fable 5.1 at 91.4%
Reasoning and knowledge
- Humanity’s Last Exam#20 of 14744.7%
Leader: Claude Opus 5.5 at 61.4%
- GPQA Diamond#3 of 14295.5%
Leader: GPT-6 Astra at 96.1%
Vision and long context
- AA Long Context Reasoning#16 of 13983%
Leader: Kimi K3 at 88.7%
Other published results
| Result | Score | Reported by | Source |
|---|---|---|---|
| DeepSWE | 70.8% | vendor | Sakana AI: Fugu Max benchmark chart ↗ |
| CharXiv Reasoning · without tools | 88.1% | vendor | Sakana AI: Fugu Max benchmark chart ↗ |
| Chartography | 37% | vendor | Sakana AI: Fugu Max benchmark chart ↗ |
| GDP.pdf | 26% | vendor | Sakana AI: Fugu Max benchmark chart ↗ |
| AutomationBench · public set | 49.5% | vendor | Sakana AI: Fugu Max benchmark chart ↗ |
| SWEFish (Sakana internal) | 62.5% | vendor | Sakana AI: Fugu Max benchmark chart ↗ |