Fugu Ultra V2
Neural Network
Fugu Ultra V2 is a Sakana AI model with multi-agent orchestration for complex composite tasks, available via BotHub API.
Specifications
- Input
- text, images, documents
- Output
- text
- Context size
- 1 000 000 tokens
- Max answer length
- 128 000 tokens
- Reasoning effort
- high, extra high, max
- Capabilities
- reasoning, function calling
Prompt cost
(per 1M tokens)
Answer cost
(per 1M tokens)
How it works Fugu Ultra V2?
Fugu Ultra v2 is the flagship model of the Fugu family from Sakana AI. Its unique design is not a single monolithic neural network, but a trained multi-agent orchestration system where a language model distributes work among agents and synthesizes their results into a final answer. Within the Fugu family, the Ultra version provides higher performance, making it ideal for multi-step tasks. In BotHub, the model is accessible without a VPN or foreign card: pay with a Russian card in rubles only for tokens actually used; they do not expire. Over 250 neural networks are available in one window: compare answers and switch models without rewriting code—the unified BotHub API remains constant, traffic is encrypted, and companies get contracts, invoices, EDI, and an admin panel with limits. Use cases: break down complex work tasks into subtasks, build an agent pipeline with multiple roles, aggregate and synthesize information from multiple sources, integrate the model into your service via API, or compare its output with other models in the catalog.Frequently asked questions about Fugu Ultra V2
You can use generated results for commercial purposes. You own all rights to the content you create. The only restriction: make sure your prompt does not include copyrighted third-party material. You are responsible for respecting the rights to any input data.
This is a text model from Sakana with a context window of up to 1,000,000 tokens, accepting both text and images/documents as input. It is suitable for working with large codebases, refactoring and debugging, analyzing reports and contracts, as well as analytics and STEM tasks.
Up to 128,000 tokens per response—enough for voluminous documentation, a large code module, or a detailed analysis of a long document. If you need a larger result, break the task into steps and pass the previous fragment into the next request via API.
Yes, reasoning before answering is confirmed: the model builds a chain of thought rather than just providing the first option that comes to mind. This is noticeable in math, logic, requirements analysis, and code debugging. In BotHub, you pay per token, so long reasoning processes don't hit subscription limits.
Yes, the model supports both function calling and structured JSON output. This is essential for agents, bots with access to your services, and pipelines where the response goes directly into code. The unified OpenAI-compatible BotHub API allows you to switch models without rewriting the integration.
Images and documents—yes: screenshots, diagrams, charts, and entire files without prior parsing into text. There is no mention of audio or video in our data, so do not count on them in advance—it is better to test the required format with your task in the interface.
The language of prompts and responses is not specified in our data, so we cannot promise specific quality for Russian. The most honest way is to run two or three of your typical queries and compare them with another model nearby: in BotHub, switching takes one click, and payment is in rubles.