DeepSeek's Liang Wenfeng: strongest models may be open-sourced, same model as deployed internally
teortaxesTex · x · 2026-10-28
teortaxesTex parses two quotes from DeepSeek founder Liang Wenfeng's investor call: "we will open-source, and even our strongest model may be open-sourced," and "we won't open-source a weaker model while deploying a better one ourselves—it will be the same model." He reads this as a costly commitment, possibly implying completed models never released or served. Follow-up: the rumored V5 code trail turned out to be September assert tests—no V5 soon. Screenshots add Liang's fuller argument: closed source offers no benefit (citing ByteDance), open source coexists with commerce if priced at 10-month payback, third parties can't match DeepSeek's deployment costs, and the company's goal is AGI.
More from Companies & People
- First-ever FastAPI Conf opens in Amsterdam with OpenAI and Pydantic talks — DynamicWebPaige · 2026-10-28
- Amodeis' Neurotech Remarks Spark Debate Over Human-AI Coevolution at Anthropic — burny_tech · 2026-10-28
- OpenAI partners with Oracle: ChatGPT, Codex and OpenAI models go native in Oracle Cloud — emmanuelvivier · 2026-10-28
- YC founder's biggest lesson: launch early — you'll never feel ready — KlausCodes · 2026-10-28
- SpaceXSI sales team builds shared Grok Bots for deals, saving hundreds of hours a month — elonmusk · 2026-10-28
- WSAI details eligibility for women's Post Bacc fellowship: 4-year degree, under 28 encouraged — ravi_iitm · 2026-10-28