Releases·Tool Releasefeatured
Z.ai open-sources ZCode after security remediation; CAICT and NSFOCUS confirm fixes in v3.14.0
Z.ai open-sourced ZCode at github.com/zai-org/ZCode in response to community-reported security issues, releasing the desktop app, web workspace, backend, Agent CLI, and runtime. The company apologized, disabled the Repo Wiki feature and local repository snapshot uploads, and published third-party security assessments from CAICT and NSFOCUS. It also committed to an ongoing vulnerability reporting process with severity-based rewards.
645 engagement1 sourceMon, Sep 21, 2026, 02:09 AM
AI Overview
- Z.ai open-sourced ZCode at github.com/zai-org/ZCode, including the desktop app, web workspace, backend, Agent CLI, and runtime.
- CAICT confirmed the zcode-prod Alibaba Cloud OSS bucket is in a zero-data state; NSFOCUS confirmed all objects and the bucket were deleted.
- ZCode v3.14.0 removes the Repo Wiki feature and disables local repository snapshot generation and upload workflows.
- Z.ai says referenced code data was never retained and never used for model training, and plans a rewarded security vulnerability reporting process.
Related
Releases·Tool Release
Developer ports SpaceXAI's pstack plugin to ZCode using GLM 5.3
A developer ported the pstack plugin, originally built by Lauren Tan at SpaceXAI for Cursor, to work with ZCode. The adaptation involved rewriting skills to fit ZCode's plugin model, using GLM 5.3 to assist the process.
A developer ported the pstack plugin, originally built by Lauren Tan at SpaceXAI for Cursor, to work with ZCode. The adaptation involved rewriting skills to fit ZCode's plugin model, using GLM 5.3 to assist the process.
13 engagement1 sourceredditThu, Aug 20, 2026, 07:00 PM
Social·Community
Users weigh GLM-5.3-Flash's agentic gains against reliability concerns at scale
A Reddit user reports GLM-5.3-Flash marks a clear step change in agentic capability but questions whether it is reliable enough to trust at scale, citing failures in the long tail of agent work. They suggest orchestrating locally served agents with stronger cloud agents as an interim approach. They also observe fewer issues in GLM's own harness ("ZCode") than in Droid, speculating the harness policies are tuned to the model.
A Reddit user reports GLM-5.3-Flash marks a clear step change in agentic capability but questions whether it is reliable enough to trust at scale, citing failures in the long tail of agent work. They suggest orchestrating locally served agents with stronger cloud agents as an interim approach. They also observe fewer issues in GLM's own harness ("ZCode") than in Droid, speculating the harness policies are tuned to the model.
43 engagement1 sourceredditSun, Aug 30, 2026, 12:34 PM
Releases·Model Releasefeatured
Z.ai releases GLM-5.3 with frontier coding and emergent cyber capabilities
Z.ai released GLM-5.3, a new model positioned as a frontier coding model with emergent cyber capabilities. The release is notable for its claimed performance in coding benchmarks and its potential security implications.
Z.ai released GLM-5.3, a new model positioned as a frontier coding model with emergent cyber capabilities. The release is notable for its claimed performance in coding benchmarks and its potential security implications.
24.3k engagement2 sourceshackernews, xFri, Aug 14, 2026, 05:19 AM
Releases·Model Releasefeatured
Z.ai releases GLM-5.3 with cyber and coding gains, holds open weights for safety review
Z.ai announced GLM-5.3, an open-weight model built on the same mixture-of-experts base as GLM-5.2, with all gains from extended post-training. It scores 84.5% on CyberGym, edging Claude Mythos 5 and GPT-5.6 Sol, and shows large jumps on Terminal-Bench 3.0 and DeepSWE v1.1. The company is holding back weights for a cyber safety review.
Z.ai announced GLM-5.3, an open-weight model built on the same mixture-of-experts base as GLM-5.2, with all gains from extended post-training. It scores 84.5% on CyberGym, edging Claude Mythos 5 and GPT-5.6 Sol, and shows large jumps on Terminal-Bench 3.0 and DeepSWE v1.1. The company is holding back weights for a cyber safety review.
Updated Tue, Aug 18, 2026, 08:46 PM — API went live on 2026-08-18, priced same as GLM-5.2, via official API and partner gateways.
1416.1k engagement6 sourcesrss, reddit, xFri, Aug 14, 2026, 12:00 AM
Releases·Model Releasefeatured
Z.ai releases GLM 5.3 Flash (Ox Alpha) with frontier intelligence at flash cost
Z.ai released GLM 5.3 Flash (Ox Alpha), a new model variant positioned as offering frontier-level intelligence at a lower cost. The release is detailed in their blog post, which includes benchmark comparisons.
Z.ai released GLM 5.3 Flash (Ox Alpha), a new model variant positioned as offering frontier-level intelligence at a lower cost. The release is detailed in their blog post, which includes benchmark comparisons.
2116.5k engagement4 sourcesreddit, xWed, Aug 26, 2026, 02:37 PM