ECNU launches Mint-Agent: Financial AI beats GPT-5.6-Sol at 22% of cost
机器之心 · wechat · 2026-08-25
East China Normal University released Mint-Agent, a family of financial-native Agentic Foundation Models (9B Mint-Cu and 27B Mint-Ag). On FinanceAgentBench v2, the 27B model scored 60.49%, surpassing GPT-5.6-Sol and Claude Opus 4.8, with per-query reasoning costs only 22% and 10% of theirs respectively. The tech focuses on auditable trajectories, binding sources, time, values, and calculations into a recoverable chain. Commercial orders have exceeded 100 million RMB.
More from Apps
- claw.cleaning: MCP Connector for Booking SF Apartment Cleanings — modelcontextprotocol · 2026-08-25
- User asks ChatGPT to roleplay as boyfriend for a week, finds it surprisingly affectionate — Ok_Novel2563 · 2026-08-25
- Anthropic Showcases Creative Claude Code Projects: From Medical Viewers to 3D Streets — 机器之心 · 2026-08-25
- Why Agentic Coding Tools Offer More Control Than Image Generators — alexisgallagher · 2026-08-25
- Postmortem of a Hindi-English voice agent for fintech: number readback and real concurrency were the real problems — admrys · 2026-08-25
- Creator uses AI to build persona, attracting female tech audience — brianryhuang · 2026-08-25