NIST puts GLM-5.3 four months behind US frontier

NIST’s CAISI assessment puts Z.ai’s GLM-5.3 about four months behind the current U.S. frontier in aggregate cyber capability, even while calling it the most capable open-weight model released to date. NIST’s benchmark report evaluates four vulnerability-discovery and exploit-development tests and says a 400-point index gap equals 10x odds of solving tasks; results are governance evidence, not a deployment recipe. Security teams should file it beside the archive’s agent-evaluation governance analysis: open weights widen access before they close the frontier gap.