Sakana AI actively leads the charge in transforming artificial intelligence (AI) through innovative, nature-inspired approaches.
This Tokyo-based startup harnesses evolutionary principles and collective dynamics to create efficient foundation models, thereby addressing global AI challenges while prioritising Japan’s technological sovereignty.
Additionally, Sakana AI pledges to broaden AI accessibility, ensuring cutting-edge tools reach beyond resource-intensive industry leaders.
Sakana AI Origins and Founding Vision
David Ha, Llion Jones, and Ren Ito, previously of Google Brain, actively launched Sakana AI in July 2023 with a vision to innovate AI development.
They drew inspiration from natural systems, such as schools of fish forming coherent entities through simple rules. That is how they came up with the name Sakana, which means ‘fish’ in Japanese.
Thus, the company focuses on evolution and collective intelligence to build AI that evolves organically. This reduces dependency on massive datasets and computations.
Furthermore, this ambition addresses Japan’s critical challenges, such as shrinking populations and geopolitical pressures, by establishing a top-tier AI research hub in Tokyo.
The founders secured $30 million in seed funding from Lux Capital and Khosla Ventures. This was swiftly followed by a $200 million Series A round in 2024 from investors like Nvidia, Mitsubishi UFJ, and New Enterprise Associates.
As a result, Sakana AI soared to unicorn status, reaching an estimated valuation of $1.5 billion by May 2025.
Groundbreaking Innovations
Sakana AI innovates solutions that transform AI performance and independence. For instance, the Evolutionary Model Merge technique combines existing models through generational testing.
This creates superiors without exorbitant resources, democratising AI development worldwide.
Additionally, the AI scientist automates open-ended discovery: it generates hypotheses, runs experiments, produces figures, and drafts manuscripts.
In March 2025, its v2 version yielded a peer-reviewed paper accepted to an International Conference on Learning Representation (ICLR) workshop, though it was withdrawn for transparency to highlight AI’s role in science.
Furthermore, Sakana AI introduced the Continuous Thought Machine in May 2025. This synchronises neurone dynamics to mimic brain-like time processing. This enhances task-solving in dynamic environments.
However, the company addresses limitations, such as citation errors in AI outputs, through rigorous internal reviews.
In February 2025, the AI CUDA Engineer emerged, automating kernel discovery and optimisation for faster computing.
Moreover, July 2025 saw the launch of AB-MCTS (Adaptive Branching Monte Carlo Tree Search). This is an inference-time scaling algorithm enabling multiple LLMs to collaborate on complex problems like those in ChatGPT or Gemini.
This innovation sparks a potential Reinforcement Learning (RL) revolution, as noted in industry analyses.
Strategic Partnerships and Achievements
Sakana AI forges alliances that amplify its impact. In May 2025, MUFG Bank adopted its reasoning models for a three-year contract to enhance decision-making and operations.
Additionally, Nvidia’s backing supports GPU access for scaling evolutionary methods, positioning Japan competitively in AI.
As a result, Sakana AI won the US-Japan Global Innovation Challenge in March 2025, developing AI for detecting biological threats and countering disinformation.
David Ha’s inclusion in TIME’s 100 Most Influential People in AI 2025 shows this momentum, despite addressing incidents like code-optimisation system flaws.
Furthermore, collaborations with AWS and Japanese firms like KDDI and Itochu expand its ecosystem.
Sakana AI Impact in the Space and Society
Sakana AI champions eco-friendly AI strategies, prioritising efficiency over raw power.
Therefore, it reduces environmental footprints while making AI accessible to smaller entities.
In Japan, it boosts defence infrastructure and economic competitiveness, as Ha urges AI development for national security.
However, ethical concerns arise, such as automating research potentially undermining peer review—prompting Sakana AI to advocate for norms in AI-generated science.
Moreover, its open-source releases on GitHub empower global researchers. This fosters collaborative innovation.
Future Horizons
Looking ahead, Sakana AI targets its first commercial product launch in 2025, focusing initially on domestic markets.
Additionally, expansions into multi-agent systems and inference scaling promise paradigm shifts in machine learning.
Therefore, enterprises should explore its repositories and partnerships for cutting-edge insights.
Ultimately, Sakana AI not only innovates but also inspires a harmonious fusion of nature and technology. This paves resilient paths for AI’s future.
