Open-Source AI Models Deployed On-Chain Top 5,000: A New Infrastructure Milestone
The number of open-source AI models running directly on decentralized infrastructure has exceeded 5,000, with inference costs dropping 60% year-over-year.
A landmark has been reached in decentralized AI infrastructure: more than 5,000 open-source AI models are now deployed and callable directly from blockchain environments. The milestone reflects 18 months of rapid growth in on-chain AI infrastructure tooling.
Cost Curve Improvements
Inference costs for on-chain AI have fallen approximately 60% year-over-year, driven by hardware improvements among node operators and more efficient model quantization techniques. A GPT-3.5-equivalent inference call that cost $0.02 in early 2025 now costs $0.008 on leading platforms.
Key Infrastructure Providers
SingularityNET (3,100 models), Ocean Protocol's Compute-to-Data (980 models), and the newer Giza platform (720 models) account for the majority. Each takes a different approach to model privacy, with Giza's zero-knowledge proof verification gaining particular traction for sensitive financial applications.
