PANews reported on May 23rd that according to the official announcement from DMind, the team has launched its first open-source large model DMind-1 for Web3 scenarios, which is fine-tuned based on Alibaba's Qwen3-32B. The model has achieved state-of-the-art performance in nine Web3 sub-tracks (such as chain infrastructure, smart contracts, DeFi, Non-Fungible Token, etc.), with inference costs only one-tenth of mainstream large models. The lightweight DMind-1-mini version retains over 95% of the performance and features lower latency. The model has been released on platforms like Hugging Face and is leading the benchmarking in the Web3 field.
