Sakana AI's Fugu-Ultra v1.1 Beats GPT-5.5 and Claude on Coding Benchmarks

Fugu-Ultra v1.1 is live. Sakana AI's orchestration-based model system just got a meaningful bump, with benchmark gains of up to 7.9 points over v1.0 and particularly strong improvements on ProgramBench and Terminal Bench 2.1. The price stays the same. If you were already using Fugu-Ultra, you're now getting a better model for free.

Not a model, an orchestra

To understand why v1.1 matters, it helps to understand what Fugu actually is. Instead of training a bigger base model, Sakana AI shipped Fugu as a system that orchestrates a pool of existing frontier models behind a single OpenAI-compatible API. Think of it as a learned conductor that reads your query and decides which combination of GPT-5.5, Claude Opus 4.8, and Gemini 3.1 Pro should handle it, and how they should collaborate.

Sakana Fugu is itself a language model trained to call various LLMs in an agent pool, including instances of itself recursively. The key insight is that different frontier models have different specializations. GPT tends to dominate in math, Gemini in scientific recall, and Opus in software engineering and debugging. Fugu learns when to call each one, and how to combine their outputs into something better than any single model could produce alone.

For v1.1, Sakana incorporated the latest frontier model releases into the agent pool. That's the core of the upgrade: as the underlying models improve, Fugu can route to better workers, and the orchestration layer learns to take advantage of their new capabilities.

What the numbers say

The headline benchmark improvements are concentrated where it matters most for real developer workflows:

• Fugu Ultra scores 82.1% on TerminalBench 2.1, ahead of GPT-5.5 at 78.2%, Claude Opus 4.8 at 74.6%, and Gemini 3.1 Pro at 70.3%. TerminalBench tests agentic coding in a live terminal environment, including file manipulation, shell commands, and multi-step execution.