Products
With an early focus on digital finance, we’ve built some high-quality open-source products: LLMs, benchmarks, datasets, and tools.
Made the first Web3 Benchmark, “DMind Web3 Benchmark”, which reached the #1 spot on Huggingface in the dataset category. It sets a new standard for evaluating LLMs on Web3-specific reasoning and domain knowledge.
Built the Web3-native language model “DMind-1” fine-tuned with expert-curated data. DMind-1 outperformed leading general-purpose models like Grok-3 and GPT-o3 on Web3-specific tasks - all while running at just 10 - 30% of their token cost.
Product Descriptions
DMind Web3 Benchmark
A holistic evaluation suite across 9 crypto-native categories (objective + subjective).
2000 expert-reviewed questions and 10+ question formats across DeFi, NFTs, DAO and more
Provides domain-level and overall scores for precise model comparison
DMind-1
A Web3-native expert LLM built on Qwen3-32B, fine-tuned on 13k+ domain Q&As and aligned via RLHF.
Optimized for expert-level reasoning in major Web3 categories like DeFi and smart contracts
Cost-efficient: 10–30% token cost of top-tier LLMs like GPT-o3
Ideal for agents, research assistants, and on-chain analytics tools
DMind-1-mini
A lighter variant (14B) retaining 95% performance with faster inference.
Excellent for real-time, low-latency applications
Cost-effective for production use at scale
Last updated