Fugu Ultra v2

sakana/fugu-ultra-v2

Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family. Rather than a single monolithic model, Fugu is a learned multi-agent orchestration system: a language model trained to...

InputFileImageText
OutputText
Input / Outputclub$5.3 / $31.8per 1M tokens · USD
Context window1Mtoken
Maximum output128Ktoken

Pricing

Current model usage rates. Choose your plan to see its published prices.

Pricing
UsagePricing
InputUS$5.3/ 1M token
InputInput length ≥ 272,001 tokensUS$10.6/ 1M token
Cache readUS$0.53/ 1M token
Cache readInput length ≥ 272,001 tokensUS$1.06/ 1M token
OutputUS$31.8/ 1M token
OutputInput length ≥ 272,001 tokensUS$47.7/ 1M token

Membership and model usage are billed separately. Compare plans

Capabilities

Input
FileImageText
Output
Text
Context window
1,000,000 token
Maximum output
128,000 token
Reasoning
max · xhigh · high

Providers

Model supply available through Vecbase.

Via OpenRouter

Access is subject to your account’s permissions and model availability.

Frequently asked questions

What is Fugu Ultra v2, and who develops it?

Fugu Ultra v2 is a model from sakana. Its model ID on Vecbase is sakana/fugu-ultra-v2. Fugu Ultra v2 is the higher-performance model in Sakana AI's Fugu family. Rather than a single monolithic model, Fugu is a learned multi-agent orchestration system: a language model trained to...

How much does Fugu Ultra v2 cost on Vecbase?

For Fugu Ultra v2, the Club plan lists USD 5.3 per 1 million input tokens and USD 31.8 per 1 million output tokens, as of 10/22/2026, 6:57:02 PM UTC. See the pricing table for cache rates, input-length tiers and other usage types. Membership fees are separate.

How many tokens fit in Fugu Ultra v2’s context window?

Fugu Ultra v2 has a context window of 1,000,000 tokens. Input and output must fit within the limits of the API you use.

What is the maximum output length for Fugu Ultra v2?

Fugu Ultra v2 can generate up to 128,000 tokens per response. Your API or request settings may set a lower limit.

Can Fugu Ultra v2 call tools in an agent workflow?

Yes. Fugu Ultra v2 supports tool calling. Your application provides the tool definitions and runs the tools the model requests.

Does Fugu Ultra v2 support structured outputs?

Yes. Fugu Ultra v2 supports structured output. Set the required format in your API request and check the returned data before using it.

Which input and output formats does Fugu Ultra v2 support?

Fugu Ultra v2 accepts File, Image, and Text and produces Text.

Which providers make Fugu Ultra v2 available through Vecbase?

Vecbase offers Fugu Ultra v2 through OpenRouter. Availability may vary by provider.

Which model ID should an agent use to request Fugu Ultra v2?

Use sakana/fugu-ultra-v2 as the model identifier for Fugu Ultra v2. See Vecbase’s API documentation for authentication, required inputs and supported request formats.

Which reasoning settings are listed for Fugu Ultra v2?

Fugu Ultra v2 supports these reasoning levels: max, xhigh, and high. Available options depend on your API.

What does an agent need to access Fugu Ultra v2 through Vecbase?

To use Fugu Ultra v2, you need an authenticated account with the required permissions, sufficient balance and the relevant capability enabled. See the documentation for setup.

Is a release date available for Fugu Ultra v2?

We do not yet have a confirmed release date for Fugu Ultra v2.

Which models can I compare with Fugu Ultra v2 on Vecbase?

You can compare Fugu Ultra v2 with Fugu Max (sakana/fugu-max), Fugu Ultra (sakana/fugu-ultra), and Sakana Namazu (sakana/sakana-namazu). Check their capabilities and prices to find the best fit for your task.