Sakana Fugu API Pricing: Plans, Costs Explained (2026)

Key Takeaways

• Sakana Fugu has two billing tracks: subscription plans ($20/$100/$200 per month) and pay-as-you-go API rates.
• All subscription tiers include both Fugu and Fugu Ultra. You do not need a higher plan to access Ultra.
• Fugu Ultra pay-as-you-go is $5 input / $30 output per 1M tokens, rising to $10/$45 above 272K context.
• Orchestration tokens are billed separately and are not included in the standard input/output count. This is the most common hidden cost that catches teams off guard.
• Pay-as-you-go users get higher processing priority than subscription plan users.
• Standard Fugu (non-Ultra) pay-as-you-go is charged at the rate of whichever top-tier model is active for that request. Sakana does not stack fees when multiple agents run simultaneously.
• There is no free tier. A free second month is available for new subscribers who sign up before July 31, 2026.

Sakana Fugu Pricing Overview

Fugu pricing looks simple on the surface: three subscription tiers and a pay-as-you-go option.

The complexity is underneath. Unlike a standard model API where one request equals one model call at a known rate, Fugu can fan out into multiple frontier model calls, verification passes, and synthesis rounds before returning one answer.

That means the surface price and the actual cost per task are two different numbers.

This guide covers both.

Subscription Plans

Subscription plans are designed for steady, predictable workloads where a fixed monthly cost matters more than exact per-token control.

Subscription Tier Breakdown

All three tiers include access to both Fugu and Fugu Ultra. Choosing a higher tier buys more capacity, not more features.

What the Tiers Do Not Tell You

Sakana does not publish exact token allowances per tier.

The public pricing page describes Pro as roughly 10x Standard usage and Max as roughly 20x, but does not commit to a specific token count. That is deliberate: Fugu’s token consumption varies depending on which underlying models activate and how much orchestration work each query requires.

Treat the plan prices as budget references, not guaranteed token equivalents.

The Free Month Offer

New subscribers who sign up before July 31, 2026, receive a free second month at their initial plan tier.

This applies to all three tiers. If you are evaluating Fugu for a production decision, this window is worth using to run a real workload rather than a demo.

Pay-As-You-Go API Pricing

Pay-as-you-go is designed for heavy production workloads that need maximum reliability and predictable per-token cost. It carries a meaningful operational advantage over subscriptions: pay-as-you-go tokens are served at higher priority than subscription plan tokens.

Fugu Ultra Rates

Fugu Ultra has fixed pricing tied to a specific model snapshot.

The current versioned snapshots are fugu-ultra-v1.1 and fugu-ultra-v1.0 (previously fugu-ultra-20260615). Both use the rates above.

Standard Fugu Rates

Standard Fugu does not have a fixed per-token rate published in advance.

You are charged the rate of whichever top-tier underlying model handled your request. Sakana does not stack fees when multiple agents are active on the same query. You pay a single rate based on the highest-tier model involved.

This means the cost of a standard Fugu call is only known after the request completes, not before.

Fugu Cyber Rates

Fugu Cyber, Sakana’s gated cyber-defense endpoint launched July 21, 2026, is available only through pay-as-you-go billing. It is not included in subscription plans.

Specific Fugu Cyber per-token rates are published on Sakana’s console pricing page. Access requires verification before billing begins.

The Hidden Cost: Orchestration Tokens

This is the most important pricing detail that most guides miss.

When Fugu Ultra processes a request, it returns usage fields that break down token consumption into two categories:

• User-visible tokens: the input you sent and the output you received
• Orchestration tokens: the internal work Fugu did coordinating agents, verifying outputs, and synthesizing results

Both are billed. Orchestration tokens are stored in token_details fields in the API response, but they represent real usage and are charged at the same input and output rates as your visible tokens.

Unlike OpenAI, even though the orchestration tokens appear in token_details fields, they represent real token usage outside of the input and output tokens and will be counted in the final price of the request.

Sakana AI pricing documentation

What This Means in Practice

A calculator that only counts your prompt and the final reply will materially understate the actual cost of a long multi-agent run.

On a complex task that triggers multiple agents and verification passes, orchestration tokens can add meaningfully to the total. Sakana does not publish average orchestration overhead per request type, so the only reliable way to know your true cost per task is to run it and inspect the full usage response.

How to Track Actual Cost

Every Fugu API response includes per-request token usage and cost reporting. Build your cost monitoring around those fields, not around an estimate based on input length and expected output length alone.

Subscription vs Pay-As-You-Go: Which Should You Choose?

The decision comes down to three factors: volume, predictability, and priority.

When Subscription Makes Sense

Choose a subscription plan if:

• Your team uses Fugu daily at a reasonably consistent volume
• You want a single line-item cost on your monthly budget
• You are still evaluating whether Fugu fits your workload and want cost certainty during the trial period

When Pay-As-You-Go Makes Sense

Choose pay-as-you-go if:

• Your usage is spiky, with long quiet periods and occasional intensive bursts
• You are running production workloads where processing priority matters
• You need access to Fugu Cyber
• Your finance team requires per-task cost attribution rather than a flat monthly allocation

Cost Comparison: Fugu vs Other AI APIs

Pricing across AI APIs is only comparable if you account for what each model actually delivers per request. A lower per-token rate from a weaker model is not cheaper if it takes three attempts to get a usable answer.

That said, a direct rate comparison is useful as a starting point.

Per-Token Rate Comparison

All rates as of July 2026. Verify current rates on each provider’s official pricing page before budgeting.

The Orchestration Premium

Fugu Ultra’s $30/1M output rate is 20% higher than Claude Opus 5’s $25/1M standard output rate. For output-heavy workloads, that gap compounds at scale.

The argument for paying the premium: Fugu Ultra coordinates multiple frontier models and runs verification passes that a single model call does not. If that orchestration meaningfully reduces errors or retries on your specific workload, the per-token premium can pay for itself.

The counter-argument: if your tasks are well-served by a single strong model, you are paying 20% more on output and absorbing orchestration token costs for coordination overhead that does not improve your outcome.

Run your own cost-per-task calculation on a representative sample before committing to either billing track.

Sample Cost Estimates

These are illustrative estimates based on published rates. Actual costs will vary based on which agents activate, orchestration overhead, and context length.

Light Use (Standard Plan: $20/month)

Suitable for fewer than ten complex queries per week. If your team is exploring Fugu for a specific use case and running occasional tests, the Standard plan covers this without needing pay-as-you-go.

Regular Professional Use (Pro Plan: $100/month)

Roughly 10x Standard capacity. Suitable for a small team using Fugu daily for code review, research tasks, or multi-step analysis. Consistent daily usage across a few people.

Production Workload Estimate (Pay-As-You-Go)

Add 20 to 40% to these estimates to account for orchestration token overhead. These are directional, not guaranteed.

API Access: How to Get Started

Getting API access takes four steps. There is no lengthy approval process for standard Fugu and Fugu Ultra.

Step 1: Create an Account

Sign up at console.sakana.ai. Standard Fugu and Fugu Ultra are available without an application.

Fugu Cyber requires a separate verification step before access is granted.

Step 2: Choose Your Billing Mode

Inside the console, set your API key’s billing mode to either subscription or pay-as-you-go. This setting determines which pricing track applies to your requests.

Step 3: Connect Your Existing Client

Fugu uses a fully OpenAI-compatible API. Set the base URL and swap your API key:

Base URL: https://api.sakana.ai/v1
Model:    fugu
          fugu-ultra-v1.1
          fugu-ultra-v1.0

No SDK migration is required. Any existing OpenAI-format client works without code changes.

Step 4: Monitor Usage Per Request

Every API response includes token usage and cost fields. Set up usage monitoring from your first request. Do not rely on estimates when orchestration tokens can shift your actual cost materially.

Geographic Availability

Fugu is not available in the EU or EEA at launch. This is the most common access blocker for international teams evaluating the API.

If your team is based in Europe or your data processing falls under GDPR, Fugu is not a current option. Check Sakana’s console for updated availability as the product expands.

Choosing the Right AI Model and Pricing Track for Your Business

The right AI model for your business is not determined by benchmark tables or per-token rates alone. It depends on what you are actually trying to do, how your team works, and what your data can support.

Phos AI Labs is an embedded AI implementation partner for mid-market companies. We are CCA-F certified by Anthropic, members of the Anthropic Claude Partner Network, and one of the first ten firms globally in the OpenAI Select Partner Network. We evaluate AI tools and pricing structures against real workloads, then build the foundations your team needs to use them well.

Whether Fugu Ultra, Claude Opus 5, or another model fits your workload best, we will tell you the honest answer before you spend a dollar on tokens.

• Model and pricing evaluation: We test your real tasks, estimate your true token costs including orchestration overhead, and recommend what fits your budget
• AI strategy: We build the decision framework before any API key is activated
• Team training: We train your team on the tools that actually fit how they work
• AI Implementation: We stay until AI is part of how the business runs, not just a line item on your cloud bill

Ready to Choose the Right AI Tools for Your Budget?

400+ engagements. Clients include Zapier, Coca-Cola, Medtronic, Dataiku, and American Express.

If you are working out whether Fugu, Claude, or another model makes economic sense for your business, start with a conversation at Phos AI Labs.

Frequently Asked Questions

How much does Sakana Fugu cost per month?

Subscription plans start at $20/month (Standard), $100/month (Pro), and $200/month (Max). All tiers include Fugu and Fugu Ultra. Pay-as-you-go is available separately at $5 input / $30 output per 1M tokens for Fugu Ultra.

Is there a free tier for Sakana Fugu?

No. There is no permanent free tier. New subscribers who sign up before July 31, 2026 receive a free second month at their initial plan tier.

What is the Sakana Fugu Ultra API pricing?

Fugu Ultra costs $5 input / $30 output / $0.50 cached input per 1M tokens, rising to $10 / $45 / $1.00 above 272K context. These rates apply to both fugu-ultra-v1.1 and fugu-ultra-v1.0.

What are orchestration tokens and are they billed?

Yes, they are billed. Orchestration tokens are the internal tokens Fugu uses to coordinate agents, verify outputs, and synthesize results. They appear in token_details fields in your API response and are charged at the same input and output rates as your visible tokens. A cost estimate that ignores them will understate your actual bill on complex multi-agent tasks.

Is pay-as-you-go or subscription better for production use?

Pay-as-you-go for production. It carries higher processing priority than subscription plan tokens and gives you exact per-request cost attribution. Subscription plans are better suited to steady daily use where a predictable monthly budget matters more than priority or granular cost tracking.

What is the Sakana Fugu API base URL?

https://api.sakana.ai/v1

Does the Pro or Max plan give access to better models?

No. All three subscription tiers include the same access to Fugu and Fugu Ultra. Higher tiers provide more capacity, not different or better models.

Can I access Fugu Cyber on a subscription plan?

No. Fugu Cyber is available only through pay-as-you-go billing and requires a separate verification process before access is granted.

Is Sakana Fugu available in Europe?

No. Fugu is not available in the EU or EEA at launch. Teams based in Europe should check Sakana’s console for updated geographic availability.

How do I calculate my expected monthly cost?

Start with your estimated request volume and average input and output token lengths. Apply the published per-token rates. Then add 20 to 40% to account for orchestration token overhead on Fugu Ultra. Run a real sample of your most common task types through the API and inspect the full usage response to get an accurate per-task cost before scaling.

All pricing figures reflect Sakana’s published rates as of July 2026 and are subject to change. Verify current rates at console.sakana.ai before making budget commitments.