TechFlow news: FLock.io has officially launched the FLock Web3 Agent Model, a large language model optimized specifically for Web3 tasks. With precise function-calling capabilities, it efficiently understands and executes complex Web3 operations. Trained by the FLock community within AI Arena, users can participate as task creators, training nodes, validators, and delegators to collectively advance the model's evolution. Built on FLock’s mainnet Task1, contributors to the model receive corresponding rewards. Compared to Web2, Web3 training resources are significantly scarcer, limiting the performance of many closed- and open-source models. The FLock Web3 Agent Model completely overcomes this bottleneck, emerging as a true Web3 intelligence expert.
The FLock Web3 Agent Model achieved a 75.93% exact match rate in Web3 function-call benchmarking, outperforming GPT-4o, Gemini Flash 2.0, DeepSeek V3, and Qwen 2.5. This milestone further validates its superior understanding, reasoning, and execution capabilities in Web3 tasks, enabling truly accurate function calling. Both the model and its evaluation framework are fully open-sourced, ensuring transparency and accessibility.
The FLock Web3 Agent Model has formed strategic partnerships with Ionet, OpenGradient, and HashKey Chain to accelerate the deployment of DeAI technology and jointly expand the application frontier of decentralized AI.
