Revision History

Auto-saved before update

Correction: reconciled the published parameter counts. Z.ai's model index labels GLM-5.3 744B-A40B, its launch post calls the base model 743B, and Hugging Face's 753,329,940,480 includes a 9,952,920,576-parameter multi-token-prediction layer (743,377,019,904 without it); added a Parameters row to the specifications table.

Update: added Artificial Analysis Intelligence Index scores for GLM-5.3 (45 on v4.3.2; 32 on v5.0, introduced October 27, 2026)

Recategorise: merge duplicate category China AI -> Chinese AI

Added related vulnerability-discovery service link

Update: evaluation wording

Correction: removed Model Context Protocol from GLM-5.3's documented features (not listed in Z.ai's model docs); noted that Z.ai labels its Anthropic comparison column both "Mythos 5" (prose) and "Fable 5 (w/ fallback)" (table, model card) instead of stating Mythos 5; updated the vLLM recipe date (September 18) and version header; removed a citation to the August 28 X post, which does not mention two additional weeks of safety work. Update: added NIST CAISI (September 17) and Anthropic Frontier Red Team (September 29) evaluations with reactions, the cvd.z.ai move to CNVD/CNNVD/NVDB, current CyberGym and ExploitGym leaderboard positions, and current Coding Plan model routing and off-peak billing.

Add section on GLM-5.3 powering Z.ai's Infra Agent that built GLM-5.3-Flash's inference service (Z.ai post, Sep 17, 2026)

Updated checkpoint availability, licensing, hosted access, and framework-specific serving requirements after the public weight release

Expanded the GLM-5.3 launch record with Z.ai's staged-release sequence, hosted-versus-checkpoint safeguard boundary, planned red-team scope, OpenVuln prototype status, and evidence limits.

No summary provided