PANews reported on May 23 that according to the official announcement of DMind, the team launched the first open source large model DMind-1 for Web3 scenarios. Based on the fine-tuning of Alibaba Qwen3-32B, it achieved SOTA performance in nine Web3 sub-tracks (such as chain foundation, smart contracts, DeFi, NFT, etc.), and the inference cost was only one-tenth of the mainstream large model. The lightweight version DMind-1-mini retains more than 95% of the performance and has lower latency. The model has been released on platforms such as Hugging Face and leads the evaluation benchmark in the Web3 field.