BREAKING
Sakana AI Launches Fugu Ultra
A 7B Conductor Calls Bigger LLMs
1
Conductor selects
↓
2
Assign roles
↓
3
Verify
↓
4
Synthesize
Fugu Ultra Benchmark Scores
GPQA-D
95.1
TerminalBench 2.1
82.1
SWE Bench Pro
73.7
SWE-Pro
54.2
Two Lines, One API Endpoint
Fugu Ultra
Quality
●
High performance
●
$5 in / $30 out per M tokens
Fugu Mini
Low latency
●
Faster responses
●
OpenAI-compatible API
Beta Praise and Skepticism
Orchestration as Next Scaling
AI NEWS BLITZ
Sakana AI has released Fugu Ultra, a system that orchestrates multiple large language models.