Sakana AI has released Fugu Max and Fugu Ultra v2, two new models in its Sakana Fugu family that act as learned orchestrators routing work across pools of other models via a single API. These releases optimize the architecture for either cost efficiency or high capability on complex tasks.
- Fugu Max targets output per dollar with pricing at $2 per 1M input tokens and $6 per 1M output tokens, claiming lower costs than Sonnet 5, GPT 5.6 Terra, and Kimi K3.
- It achieves the best overall score on six benchmarks including Terminal Bench 2.1 and GPQA Diamond by routing tasks to the leanest capable models.
- Fugu Ultra v2 targets complex reasoning and software development, scoring 48.3 on Chartography and 74.3 on DeepSWE without relying on Fable 5 or GPT-6-Astra.
- Both models are available today via Sakana’s OpenAI-compatible API, allowing users to switch with a single line of code.
The release aims to reduce vendor lock-in and exposure to API revocations by providing frontier output capabilities independent of any single proprietary model.