-
All is Not Lost: LLM Recovery without Checkpoints
Paper • 2506.15461 • Published • 37 -
NoLoCo: No-all-reduce Low Communication Training Method for Large Models
Paper • 2506.10911 • Published • 8 -
Verde: Verification via Refereed Delegation for Machine Learning Programs
Paper • 2502.19405 • Published • 2 -
HDEE: Heterogeneous Domain Expert Ensemble
Paper • 2502.19385 • Published • 2
AI & ML interests
We network together the core resource for machine intelligence to flourish alongside human intelligence https://www.gensyn.ai/research
Organization Card
The network for machine intelligence
We network together the core resources required for machine intelligence to flourish alongside human intelligence.
- Consistent Execution, ensuring compatibility across any device;
- Trustless Verification, checking and agreeing on work performed in a scalable way;
- Efficient Communication, sharing workloads between devices over the internet; and
- Decentralised Coordination, identifying participants, aligning incentives, and managing attribution.
The Gensyn Protocol brings these components together into a single network for open machine learning - accessible to humans and machines alike and scalable to every computer in the world.
Get involved
- Swarm Participants: Run a node, improve the overall intelligence of the swarm, track your participation, and train your own local model to use as you wish.
- Developers: Train a local model by participating in the swarm, deploy in local applications without being beholden to an external company, and continually update its parameters in a living, decentralised system.
- ML Researchers: Deploy your own swarm for others to join, train on new datasets, solve new problems, construct new objectives, incentivise participation and explore the space of decentralised AI with an entirely new infrastructure stack.
- Community Members: Stay up to date on progress, provide feedback, and discuss future developments.
RL Swarm is an open source system for peer-to-peer gossip-based reinforcement learning over the internet.
-
Gensyn/Qwen2.5-0.5B-Instruct
Text Generation • 0.5B • Updated • 4.02M • 20 -
Gensyn/Qwen2.5-1.5B-Instruct
Text Generation • 2B • Updated • 57.5k • • 13 -
Gensyn/Qwen2.5-32B-Instruct-bnb-4bit
Text Generation • 18B • Updated • 338 -
Gensyn/Qwen2.5-72B-Instruct-bnb-4bit
Text Generation • 39B • Updated • 114
-
All is Not Lost: LLM Recovery without Checkpoints
Paper • 2506.15461 • Published • 37 -
NoLoCo: No-all-reduce Low Communication Training Method for Large Models
Paper • 2506.10911 • Published • 8 -
Verde: Verification via Refereed Delegation for Machine Learning Programs
Paper • 2502.19405 • Published • 2 -
HDEE: Heterogeneous Domain Expert Ensemble
Paper • 2502.19385 • Published • 2
RL Swarm is an open source system for peer-to-peer gossip-based reinforcement learning over the internet.
-
Gensyn/Qwen2.5-0.5B-Instruct
Text Generation • 0.5B • Updated • 4.02M • 20 -
Gensyn/Qwen2.5-1.5B-Instruct
Text Generation • 2B • Updated • 57.5k • • 13 -
Gensyn/Qwen2.5-32B-Instruct-bnb-4bit
Text Generation • 18B • Updated • 338 -
Gensyn/Qwen2.5-72B-Instruct-bnb-4bit
Text Generation • 39B • Updated • 114
models
5

Gensyn/Qwen2.5-7B-Instruct
Text Generation
•
8B
•
Updated
•
1.23k
•
•
3

Gensyn/Qwen2.5-72B-Instruct-bnb-4bit
Text Generation
•
39B
•
Updated
•
114

Gensyn/Qwen2.5-32B-Instruct-bnb-4bit
Text Generation
•
18B
•
Updated
•
338

Gensyn/Qwen2.5-1.5B-Instruct
Text Generation
•
2B
•
Updated
•
57.5k
•
•
13

Gensyn/Qwen2.5-0.5B-Instruct
Text Generation
•
0.5B
•
Updated
•
4.02M
•
20
datasets
0
None public yet