Sakana AI Launches Marlin, an 'Ultra Deep Research' Agent for Enterprise Strategy
Tokyo-based AI startup Sakana AI has introduced its first commercial product, Sakana Marlin, an autonomous B2B research agent designed for in-depth, long-horizon reasoning. Marlin functions as a "Virtual CSO" capable of generating comprehensive, 100-page strategy reports and executive slides within an eight-hour timeframe. Unlike conventional chatbots that prioritize speed, Marlin focuses on methodical reasoning, data gathering, cross-referencing, and hypothesis formulation for corporate, financial, and think tank clients. The platform relies on advanced AI models, including its proprietary Adaptive Branching Monte Carlo Tree Search (AB-MCTS) technology, to deliver deeply researched insights.
Sakana AI, a Tokyo-based artificial intelligence startup, has officially launched its inaugural commercial product, Sakana Marlin. Billed as a "Virtual CSO" (Chief Strategy Officer), Marlin is an autonomous business-to-business research agent engineered for deep, long-horizon reasoning, diverging from the instantaneous text generation of typical chatbots.
This new AI tool can execute continuous, self-governing reasoning loops for up to eight hours. Its output includes deeply researched, well-cited strategy reports of over 100 pages, complete with executive slides. The platform is designed exclusively for enterprise use, targeting corporations, financial institutions, and think tanks.
The workflow for Marlin involves the user providing a core research topic. Following an initial exchange to refine the investigation's scope, the system operates autonomously. It formulates hypotheses, navigates the web for data, cross-references sources, and maps causal dynamics within complex business environments. The final output provides a structured set of strategic options, including executive summary slides, appendices, references, and a comprehensive report.
Marlin's underlying technology is based on Sakana AI’s Adaptive Branching Monte Carlo Tree Search (AB-MCTS) and frameworks derived from "The AI Scientist" research project. This exploration engine leverages a Multi-LLM AB-MCTS architecture, enabling it to dynamically select and coordinate various AI models for specific sub-tasks, ensuring rigorous verification and error correction throughout the research process. The AB-MCTS framework was initially introduced in June 2025 and released as the open-source library TreeQuest.
Sakana AI emphasizes a strict, enterprise-grade data policy, ensuring customer data and inputs are not used for model training or fine-tuning without explicit opt-in consent. Even with consent, data undergoes processing to remove personally identifiable information. The company, formed in 2023, was co-founded by Llion Jones, known for co-authoring Google’s "Attention Is All You Need" paper, and David Ha, a former Google Brain researcher.
Sakana AI secured a Series B funding round in late 2025, pushing its post-money valuation past $2.6 billion. Key investors include Khosla Ventures, Lux Capital, New Enterprise Associates (NEA), Nvidia, Google, Mitsubishi UFJ Financial Group (MUFG), Citi, and Salesforce. A closed beta test, which commenced in April 2026 with approximately 300 professionals, indicated positive feedback regarding Marlin's autonomous and fact-driven approach. (Source: VentureBeat)
