Is Sakana AI's Fugu better than Anthropic's Claude Fable 5?
What's the story
Tokyo-based artificial intelligence (AI) start-up Sakana AI has launched a new platform called Fugu. The company claims that Fugu can outperform Anthropic's Claude Fable 5 in certain coding benchmarks. Unlike traditional AI systems that rely on a single large language model, Fugu is designed to coordinate multiple models through a unified interface for complex tasks.
Benchmark results
Fugu outperforms Claude on coding benchmark
Sakana AI shared benchmark results showing that its Fugu Ultra model scored 93.2 on LiveCodeBench, while the standard Fugu model scored 92.9. This is better than Anthropic's Claude Fable 5, which scored 89.8 on the same benchmark. LiveCodeBench is an open-source evaluation framework that tests coding and software problem-solving skills with regularly updated programming challenges.
GPQA-Diamond performance
Fugu excels in scientific reasoning tasks
Sakana AI also reported strong results on GPQA-Diamond, a benchmark featuring graduate-level multiple-choice questions in biology, chemistry, and physics. Both Fugu and Fugu Ultra scored 95.5, surpassing the 94.6 score of Anthropic's earlier Claude Mythos Preview model. These results show that Sakana AI's platform can handle advanced scientific reasoning tasks as well as coding workloads.
Versatility
Two versions of Fugu launched
Sakana AI has launched two versions of its platform: the standard Fugu model for coding, chat, and general-purpose AI tasks, and the more powerful Fugu Ultra for demanding workloads such as AI research, paper reproduction, cybersecurity analysis, and patent investigations. Internal tests showed that the platform outperformed models like Google's Gemini 3.1 Pro, OpenAI's GPT-5.5, and Anthropic's Opus 4.8 in several specialized tasks.
Founders
Sakana AI challenges traditional AI development methods
Founded in 2023 by Llion Jones, a co-author of Google's "Attention Is All You Need" paper, and David Ha, former head of research at Stability AI, Sakana AI hopes that combining multiple AI systems can provide a scalable way to improve performance. As competition in the industry increasingly focuses on coding, reasoning, and research capabilities, Sakana's multi-model strategy offers an alternative to the race for larger foundation models.