Luo Fuli: AGI Has Already Been Achieved — the Next Step Is "Self‑Evolution", as Zhipu Rolls Out GLM‑5.1

Claim and context

It has been reported that Luo Fuli said artificial general intelligence (AGI) has already been achieved and that the industry's next phase will be "self‑evolution." That is a bold provocation in a debate still defined by contested definitions and benchmarks. Reportedly, Luo framed the announcement not as an endpoint but as the start of systems adapting and improving themselves without human‑curated updates. For Western readers: AGI remains controversial — most researchers distinguish narrow, highly capable models from the general, autonomous intelligence Luo describes.

Immediate product news: GLM‑5.1 is live

Separately, Zhipu AI (智谱) has officially published GLM‑5.1, and it has been reported that all Coding Plan tiers (Max, Pro and Lite) can now invoke the model. Practical deployment notes have circulated for users integrating GLM‑5.1 into Claude Code and OpenClaw: change a few Anthropic environment defaults, append a GLM‑5.1 entry to the models array, update the default primary model to zai/glm‑5.1, and restart the OpenClaw gateway. The configuration steps are technical but straightforward for developers; they also reveal GLM‑5.1’s aggressive context window and token limits aimed at long‑form reasoning.

Why this matters — technology and geopolitics

Why should outsiders care? Because advanced Chinese models such as GLM‑5.1 are being positioned as home‑grown alternatives to Western systems at a time of intensified technology competition. Sanctions and export controls on chips and AI tooling have pushed many Chinese teams to optimize software and system design around domestic compute. If Luo’s AGI claim is to be taken seriously, the debate will quickly move from benchmarks to governance: who controls self‑evolving systems, under what rules, and how will cross‑border risk be managed?

The unanswered questions

Reportedly, Luo’s next‑step rhetoric — "self‑evolution" — raises immediate technical and ethical questions. How would such evolution be constrained? Who audits it? For now, GLM‑5.1’s release is a concrete milestone developers can test. Luo’s larger claim remains to be verified by independent evaluation and peer scrutiny. Will the field agree that AGI is here, or will this be another milestone in the long arc of incremental progress? Only rigorous, transparent testing will tell.