Sakana: Fugu Max by Kilo Gateway | AI model information

$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3$10 off the fastest DeepSeek V4.1 Flash, Kimi K3 and GLM 5.3

Model details

Sakana: Fugu Max

Sakana AI describes Fugu Max as an orchestration system rather than a single monolithic foundation model, built around a learned model coordination architecture that the company has detailed in ICLR 2026 work. The system draws on Sakana's largest pool of open and specialized underlying models to date, dynamically routing each request to the most appropriate component. This positioning reframes the competitive frontier as a two-dimensional tradeoff between capability and cost, with Fugu Max optimized to deliver the best possible output at the lowest viable spend on each task.

In practical terms, Fugu Max is aimed at builders and enterprises who want strong reasoning and tool-using behavior without paying the premium of the most expensive proprietary models. It launched alongside the higher-capability Fugu Ultra v2 sibling, which targets maximum performance on complex multi-step problems, giving users a coordinated pair that shares the same orchestration backbone. The system's value proposition is decoupling performance from proprietary lock-in: by composing specialized models intelligently, Fugu Max is positioned for workflows that span diverse task types and benefit from adaptive, cost-aware routing.

Kilo Gatewaysakana/fugu-maxfugu

Open provider page Provider docs

Quick Info

Provider: Kilo Gateway
Model key: sakana/fugu-max
Release date: Sep 11, 2026
Last updated: Sep 11, 2026
Input modalities
Output modalities
Capabilities

Cost

Input token cost: $2.00
Output token cost: $6.00

Limits

Output tokens: 128,000 tokens
Context window: 1,000,000 tokens

Transparent token rates

Compare Sakana: Fugu Max pricing

Rates are shown per one million tokens. Combined means one million input plus one million output tokens.

Browse this family

Latest news about Sakana: Fugu Max

Kilo Gateway

Coverage

The AI Weekly news digest entry for September 11, 2026 reports that "Sakana AI released Fugu Max and Fugu Ultra v2 on September 11 as a single learned orchestrator that routes queries across a pool of open-weight and specialist models behind one OpenAI-compatible API." It names the September 11 release date, attributes As a daily aggregator brief, the entry corroborates the release timing, Sakana authorship, and headline pricing of Fugu Max but adds limited independent technical depth beyond those confirmation points. The competitive comparison ("40-60% below Sonnet 5 and GPT 5.6") is presented without a primary citation in the excer

Kilo Gateway

CoverageBenchmark

A review published September 11, 2026 explicitly covers "Fugu Max & Fugu Ultra v2" as two operating points of Sakana's learned multi-agent orchestration architecture, with Fugu Max positioned as the cost-performance model. It states Sakana released both on September 11, 2026 and lists Fugu Max at $2 per million input t On performance, the review attributes benchmark claims directly to Sakana rather than to independent reproduction, reporting best-overall results on six benchmarks for Fugu Max (Terminal Bench 2.1, GPQAD, AA-LCR, GDP.pdf, AutomationBench, and SWEFish) and stating that Fugu Max "expands the cost-performance Pareto front

Kilo Gateway

CoverageBenchmark

The OpenRouter models directory lists "Sakana: Fugu Max" as an entry published by "sakana" and dated September 11, 2026, explicitly matching the subject's exact variant name (sakana/fugu-max). The entry describes Fugu Max as the cost-performance operating point of Sakana AI's Fugu family, where "Fugu is a learned multi The same OpenRouter listing distinguishes Fugu Max from the higher-performance Fugu Ultra v2 sibling, confirming Fugu Max as a separate variant rather than a naming alias or gateway repackaging. It preserves explicit creator attribution to Sakana rather than to OpenRouter or the Kilo Gateway, satisfying the model-focus