Sakana AI's orchestration API that routes tasks across a pool of LLMs through one OpenAI-compatible endpoint.
Updated EU status checked Last shown
Sources: the vendor's own pages (linked above) and creator videos at the linked timestamps. EU status is information, not legal advice; see how we check.
Sakana Fugu is a multi-agent system delivered as a single model API. Rather than being a single base model, it dynamically assembles and coordinates a pool of underlying models, selecting and switching between them per task. Sakana says it is grounded in two ICLR 2026 papers on learned orchestration, TRINITY (an evolved coordinator assigning Thinker, Worker and Verifier roles) and the Conductor (a reinforcement-learning-trained coordinator that designs natural-language coordination strategies).
It comes in four variants: Fugu (default, balancing performance and latency), Fugu Ultra (a deeper pool for hard, high-stakes problems such as Kaggle competitions, paper reproduction and patent research), Fugu Max (the largest pool, aimed at cost-performance efficiency) and Fugu Cyber (specialised for cybersecurity reasoning). Users can opt specific providers or models out of the pool to meet data, privacy or compliance requirements. It can be plugged into tools such as Codex for coding and code review.
Access is through a paid, closed, OpenAI-compatible API. Sakana reports that Fugu Ultra (v2.0) achieves the best or joint-best score on five of eight benchmarks. Reviewers noted that, as an ensemble router, comparisons with single frontier models are not like-for-like, and one tester found results somewhat weaker than a leading frontier model on coding tasks at about $30 of usage. The service is currently not offered in the EU/EEA.
Watch the moment: the exact second top creators show it. Videos load only when you press play, via youtube-nocookie.com.
Paid, closed orchestrator that routes prompts across closed and open models, claimed to beat frontier models and match Fable 5 and Mythos preview. The creator calls the claim misleading since it's an ensemble router rather than a base model, and criticizes the pricing and lack of transparency.
An orchestrator model that routes prompts to multiple underlying models, with Fugu and Fugu Ultra variants. Benchmarks match or beat Fable 5 and Mythos on some tests. The creator tested it via Codex CLI to clone Megabonk and the Future Tools site, which cost about $30 and gave results somewhat weaker than Fable.
Closest neighbours by meaning, not just by category.
Model · NVIDIA
NVIDIA's open-weight ~550B mixture-of-experts LLM with a hybrid Mamba-transformer design, aimed at agentic workloads.
Model · Alibaba
Alibaba's upgraded Qwen3.8-Max snapshot with 1M context, thinking mode and stronger coding and agent performance, via API.
Model · Zhipu AI
Zhipu AI's GLM 5.2 large language model family, with Fast and max variants, available via Z.ai chat and API.
Model
Open 7.39B dense model for math reasoning and agentic search, with a 256K context and the full training pipeline released.
Model · OpenAI
OpenAI frontier language model used for coding, reasoning and agentic tasks, available via ChatGPT and the API.
Model · Zhipu AI
Zhipu AI's GLM-5.3-Flash, a fast open-weights LLM for coding and agentic tasks, available via Z.ai chat and API.
