Sakana AI Launches Biology Foundation Model

Sakana AI, the Tokyo-based artificial intelligence startup backed by SoftBank and founded by former Google researchers, has launched a new foundation model specifically designed for biological research. The model represents a significant push into the intersection of AI and life sciences, an area attracting billions of dollars in investment from both tech giants and pharmaceutical companies.

The move positions Sakana AI as a direct competitor to Western biotech-AI players like Google DeepMind, which made headlines with AlphaFold, and startups such as Recursion Pharmaceuticals and Isomorphic Labs. It also signals that Japan's AI ecosystem is maturing beyond general-purpose language models into specialized, high-value scientific domains.

Key Takeaways

• Sakana AI has launched a biology-focused foundation model targeting drug discovery and molecular biology applications
• The startup has raised over $300 million in funding, with SoftBank as a lead investor
• Co-founders include Llion Jones, a co-author of the landmark 'Attention Is All You Need' transformer paper, and David Ha, a former Google Brain researcher
• The model is designed to process and reason across multiple biological data modalities, including protein sequences, gene expression data, and molecular structures
• Sakana AI's approach leverages its proprietary 'nature-inspired' AI techniques, differentiating it from conventional large language model architectures
• The launch comes amid a global race to apply AI to accelerate pharmaceutical R&D timelines

Sakana AI Enters the BioAI Arena With Specialized Model

Sakana AI's new biology model is not simply a large language model fine-tuned on scientific papers. Instead, the company says it has built a multimodal foundation model from the ground up to understand biological systems at multiple scales — from individual amino acid sequences to complex cellular pathways.

This approach contrasts with efforts by companies like Meta, which released its ESMFold protein prediction tool, or even Google DeepMind's AlphaFold 3, which focuses primarily on protein structure prediction. Sakana AI's model reportedly aims to be more general-purpose within biology, capable of handling tasks across genomics, proteomics, and molecular interaction prediction.

The company has not disclosed the exact parameter count or training dataset size. However, sources familiar with the project indicate the model was trained on a curated corpus of biological databases, published research, and proprietary experimental data obtained through partnerships with Japanese pharmaceutical firms and academic institutions.

Why Biology Is the Next Frontier for Foundation Models

The pharmaceutical industry spends an estimated $2.6 billion on average to bring a single drug to market, with development timelines often stretching beyond 10 years. AI promises to compress both the cost and the timeline by identifying promising drug candidates faster, predicting molecular behavior more accurately, and reducing the number of failed clinical trials.

Google DeepMind's AlphaFold has already demonstrated the transformative potential of AI in biology, predicting the 3D structures of over 200 million proteins. But protein structure is just one piece of the puzzle. Understanding how proteins interact, how genes are expressed in disease states, and how small molecules bind to biological targets requires a more comprehensive computational approach.

Sakana AI appears to be betting that a single, unified foundation model capable of reasoning across these diverse biological data types will be more powerful than a collection of specialized tools. This mirrors the trend in natural language processing, where general-purpose models like GPT-4 and Claude have proven more versatile than task-specific predecessors.

Sakana AI's Nature-Inspired Approach Sets It Apart

What makes Sakana AI distinctive in the crowded AI landscape is its 'nature-inspired' methodology. The company's name itself — 'sakana' means 'fish' in Japanese — reflects its philosophy of drawing on evolutionary and biological principles to design AI systems.

Rather than relying solely on scaling transformer architectures with more parameters and compute, Sakana AI has explored techniques like:

• Evolutionary algorithms that mimic natural selection to optimize model architectures
• Swarm intelligence approaches inspired by collective behavior in biological systems
• Model merging techniques that combine multiple pre-trained models into a single, more capable system
• Neural architecture search guided by biological principles of efficiency and adaptation

This philosophy has already yielded results. Earlier in 2024, Sakana AI published research on its model merging technique, which demonstrated that combining existing open-source models using evolutionary strategies could produce new models that outperformed their individual components — without requiring expensive retraining from scratch.

Applying these same principles to biological AI could give the company an edge. Biological systems are inherently complex, hierarchical, and evolved — making nature-inspired computational approaches a natural fit for modeling them.

SoftBank's $300 Million Bet on Japanese AI Leadership

Sakana AI's biology model launch comes on the heels of a massive fundraising trajectory. The company has secured over $300 million in total funding, with its most recent round valuing the startup at approximately $1.5 billion. SoftBank has been a prominent backer, consistent with CEO Masayoshi Son's vocal commitment to making Japan a global AI powerhouse.

The investment landscape for AI-biology startups has been heating up globally:

• Recursion Pharmaceuticals (US) has a market cap exceeding $3 billion and partners with NVIDIA on biological AI
• Isomorphic Labs (UK), a Google DeepMind spinoff, is working on AI-driven drug discovery
• Insilico Medicine (Hong Kong/US) has advanced AI-discovered drug candidates into clinical trials
• Xaira Therapeutics (US) launched in 2024 with over $1 billion in funding for AI-driven drug design
• BioNTech (Germany) has been acquiring AI companies to bolster its mRNA platform

Sakana AI's entry into this space adds a significant Japanese contender. Japan's pharmaceutical industry is the 3rd largest in the world, and the country's academic research institutions produce world-class biological data. Sakana AI is well-positioned to leverage these domestic advantages while competing on a global stage.

What This Means for Developers and Researchers

For the developer and research communities, Sakana AI's biology model could open several doors. If the company follows its previous pattern of releasing open-weight models or providing API access, it would give computational biologists and bioinformatics researchers a powerful new tool.

Practical applications could include:

• Drug target identification: Using the model to predict which proteins or genes are most likely to be effective targets for new therapies
• Molecular generation: Designing novel small molecules or biologics optimized for specific therapeutic targets
• Literature synthesis: Automatically extracting and connecting insights from millions of biomedical research papers
• Clinical trial optimization: Predicting patient responses and identifying biomarkers for stratification

However, challenges remain. Biological AI models require extensive validation against experimental data before they can be trusted in clinical settings. Regulatory frameworks for AI-assisted drug discovery are still evolving, particularly in the US (FDA) and Europe (EMA). Sakana AI will need to demonstrate not just computational performance on benchmarks, but real-world biological accuracy.

Looking Ahead: The Convergence of AI and Life Sciences Accelerates

Sakana AI's biology foundation model launch is part of a broader acceleration in the convergence of artificial intelligence and life sciences. The global AI in drug discovery market is projected to reach $11.2 billion by 2028, growing at a compound annual rate of over 30%.

Several trends will shape this space in the coming 12 to 18 months. First, expect more foundation model providers — including OpenAI, Anthropic, and Google — to announce biology-specific models or partnerships with pharmaceutical companies. Second, the availability of high-quality biological training data will become a key competitive differentiator, potentially giving companies with strong academic and industry partnerships a structural advantage.

Third, regulatory clarity will be critical. The FDA has already begun issuing guidance on AI in clinical settings, and the European Medicines Agency is developing its own frameworks. Companies that proactively engage with regulators will be better positioned for commercialization.

For Sakana AI specifically, the next milestones to watch include potential partnerships with major pharmaceutical companies, publication of peer-reviewed validation studies, and any announcements regarding API access or open-source releases. The company's unique nature-inspired approach, combined with SoftBank's deep pockets and Japan's pharmaceutical ecosystem, gives it a compelling — if still unproven — thesis in one of AI's most consequential application domains.

The race to build AI that truly understands biology is still in its early innings. But with well-funded players like Sakana AI entering the field alongside DeepMind and a growing cohort of biotech-AI startups, the pace of innovation is accelerating rapidly.