Quant-Arena: 9 LLM agents trade real A-shares, only 4 profit and none beat equal-weight
mhdfaran · x · 2026-10-22
Einsia AI released Quant-Arena, an end-to-end quantitative research benchmark on real A-share market data testing whether LLM agents can build signals from raw data and profit on unseen years.
- 9 frontier models, 3,042 stocks; research feedback from 2022–2023, sealed test year 2024
- Only 4 of 9 models made money; top scorer GPT-6 Astra returned 10.7% in 2024
- Yet the equal-weight benchmark across all 3,042 stocks returned 17.1% — even the best model lost to the dumbest baseline
- Evaluation includes trading costs and estimated market impact
A useful test for agent builders: not whether the code runs, but whether strategies beat a baseline outside their research window. The arena accepts submissions from any model or agent harness.
Related event: Quant-Arena Benchmark: Only 4 of 9 LLMs Profit Trading A-Shares(2 posts)→
More from coding & agent
- Dev builds an agent org chart in Copilot where each specialist keeps its own second brain — PaulShellDev · 2026-10-22
- Email can poison your AI assistant's recommendations — give approvers one source the agent can't touch — YvesMulkers · 2026-10-22
- Steve Yegge's agent autonomously researches a cert vulnerability via TheHackersNews — Steve_Yegge · 2026-10-22
- Free "AI University" on GitHub: 7 weeks from Docker to agentic RAG — mdancho84 · 2026-10-22
- Entoli: open-source cross-platform agent platform with your own keys, planning and P2P sync — ProgrammingDev · 2026-10-22
- 1Password launches Agents platform: login without ever showing passwords — dotey · 2026-10-22