Zoo Labs Foundation
Open AI Research
Building the Zen family of frontier AI models. 45+ open-source models from 600M to 480B parameters. Formal verification. Community governed. Safe by design.
The non-profit open-source lab behind the Zen family of AI models.
480B
Parameters in our largest open Zen model—community-trained, verifiable weights, no corporate gatekeepers.
Explore Models25+
Zoo Improvement Proposals (ZIPs) governing research direction, model architecture, and funding allocation.
View ZIPs600+
Open source repositories across Hanzo, Lux, and Zoo powering the decentralized AI stack.
Open Source99.8%
Cost reduction in AI training via Training-Free GRPO — $18 vs $10,000+ for traditional fine-tuning.
Read PaperThe principles that drive us.
We build open AI models for everyone.
We publish research without paywalls.
We verify with formal proofs.
We train AI that is safe for children.
We govern through community proposals.
Products
Open-source tools for AI research, training, and deployment
Zoo Agents
AI-powered digital creatures. The original NFT animals, now autonomous agents with memory, goals, and on-chain identity.
Learn more →Zoo AI
Desktop app for local AI inference. Run Zen models on your hardware — free, open-source, private by default.
Learn more →Zoo Gym
Open-source training platform. Fine-tune any model with LoRA, GRPO, DPO, and 8 more methods.
Learn more →Zoo Exchange
Decentralized exchange for $ZOO and $AI tokens. Trade, stake, and participate in governance.
Learn more →Collections
Curated NFT collections of AI-generated endangered species art. Each purchase funds conservation research.
Learn more →The Zen Model Family
Open-source frontier models from edge to cloud
Foundation
zen4-nano (0.6B) to zen4-ultra (480B MoE)
General purpose language models
Code
zen4-coder-flash to zen4-coder-pro
59.2% SWE-bench verified
Multimodal
zen-omni, zen-vl, zen-3d, zen-director
Text, vision, audio, video, 3D
Safety
zen-guard family
Content safety and alignment
Zoo Gym
Open-source AI training platform
Train, fine-tune, and align any model. LoRA, GRPO, DPO, QLoRA — 8 training methods, 100+ model support, multi-GPU distributed training.
Training-Free GRPO
99.8% cost reduction. $18 vs $10,000+
Decentralized Training
DSO protocol for community-coordinated optimization
Formal Verification
15 Lean 4 proofs. Every critical system verified.
Research
7 research papers. 102 ZIP specifications. 15 formal proofs.
HLLM with Training-Free GRPO
99.8% cost reduction, 82.7% AIME24 accuracy
Experience Ledger & DSO
Byzantine-robust decentralized semantic optimization
Zoo Network Architecture
L2 AI specialization with federated learning
Zoo Network
Decentralized AI compute with recursive self-learning
Contribute GPU compute, earn $AI tokens. Agentic systems that recursively improve through the Experience Ledger. Post-quantum secure. FHE for private inference.
Leadership
Advancing open AI research since 2021
Antje Worring
Executive Director & Co-Founder
Building safer AI for children since founding Zoo Labs Foundation in 2021.
Zach Kelling
Co-Founder & CTO
Technical founder. Leads architecture across Zen models, Gym, and Zoo Network.
Woo Bin
VP Engineering
Full-stack and AI engineer leading Zoo AI, Gym, and desktop app development.
Marcus White
VP Research
Advancing Zen model capabilities. Applied AI research to production.
Dave Lorenzini
Chief Strategy Officer
Decades of experience in immersive computing and AI strategy.
Major "Dream" Williams
Chief Visionary Officer
International collaboration to transform AI for social good.
Danielle Savage
Chief Brand Officer
Elevating Zoo's global presence and research communications.
Anastasia Zacharaoff
VP Engineering
Building high-performing engineering teams and scalable AI systems.
Jackson Mori
VP Engineering
Distributed systems, performance, reliability, and developer experience.