# Z.ai Published articles for Z.ai. This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Deep Principle Releases Hydrogen-1.0: A 753B Open-Weight Model for Scientific Discovery DevFeed: [Deep Principle Releases Hydrogen-1.0: A 753B Open-Weight Model for Scientific Discovery]() Original publisher: [Read original article]() Author: Asif Razzaq Published: 2026-11-02T18:41:41Z Content type: release Language: en Sources: [MarkTechPost]() Topics: [AI for science](), [post-training](), [Fine-tuning](), [quantization](), [LLM optimization](), [Artificial Intelligence]() Tags: [agent](), [agent-discovery](), [agentic-ai](), [ai-shorts](), [applications](), [artificial-intelligence](), [chemistry](), [discovery](), [editors-pick](), [for-developers](), [for-devs](), [glm-5-3](), [hydrogen](), [laboratory](), [language-model](), [large-language-model](), [license](), [life-sciences](), [machine-learning](), [materials-science](), [model](), [new-releases](), [open-source](), [open-weights](), [post-training](), [records](), [science](), [staff](), [tech-news](), [technology](), [z-ai]() ### AI overview Hydrogen-1.0 is a 753B-parameter, MIT-licensed scientific model made by post-training Z.ai's GLM-5.3 with domain data, laboratory records and discovery trajectories. Deep Principle reports 38.3% on FrontierScience-Research versus GPT-6-Astra's 33.3%, and 93.0% on GPQA-Diamond, roughly matching Kimi-K3. Results vary by task: Hydrogen-1.0 leads GPT-6-Astra on scientific computing and forward reaction prediction, while trailing on NMR structure elucidation. Demonstrations include molecular simulation, reaction-network growth, NMR candidate ranking and experiment-protocol generation. The excerpt says serving was tested with SGLang across eight GPUs; NVFP4 and MXFP4 builds target specified NVIDIA and AMD accelerators. Full training datasets and recipes are not released, and the reported benchmark results are attributed to Deep Principle. ### Source excerpt Deep Principle has released Hydrogen-1.0, an open-weight scientific model for chemistry, materials science and life sciences. It is built by post-training Z.ai's GLM-5.3 on domain data, laboratory records and agent discovery trajectories. For developers, it offers open weights that report a FrontierScience-Research win over GPT-6-Astra. The weights ship under the MIT License. TL;DR What is [...] The post Deep Principle Releases Hydrogen-1.0: A 753B Open-Weight Model for Scientific Discovery appeared first on MarkTechPost. ## Turning a 2012 Kindle into an AI quota dashboard DevFeed: [Turning a 2012 Kindle into an AI quota dashboard]() Original publisher: [Read original article]() Author: Ivan Morgillo Published: 2026-10-13T10:00:00Z Content type: article Language: en Sources: [Ivan Morgillo]() Topics: [Plover](), [Claude](), [ClojureScript]() Tags: [ai](), [claude](), [claude-code](), [codex](), [codex-cli](), [grok](), [kindle](), [rust](), [z-ai]() ### AI overview A Rust program on a PC turns a 2012 Kindle Paperwhite into a display for remaining Claude, Codex, Z.AI and Grok subscription quota. The article explains jailbreaking the Kindle, reading usage endpoints with existing CLI credentials, rendering a grayscale dashboard, and updating it over USB every ten minutes. It notes that the endpoints are undocumented and that token refreshes are left to the provider CLIs. ### Source excerpt My 2012 Kindle Paperwhite shows how much Claude, Codex, Z.AI and Grok quota I have left. A Rust program on my PC redraws it over USB every ten minutes. ## Can an Open Model Do Security Research? Cantina's apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks DevFeed: [Can an Open Model Do Security Research? Cantina's apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks]() Original publisher: [Read original article]() Author: Michal Sutter Published: 2026-10-05T01:47:28Z Content type: news Language: en Sources: [MarkTechPost]() Topics: [Security research](), [benchmark overfitting machine learning](), [rlvr](), [llama]() Tags: [agent-harness](), [agentic-ai](), [ai-shorts](), [applications](), [artificial-intelligence](), [benchmark](), [deep-learning](), [editors-pick](), [glm-5-3](), [language-model](), [large-language-model](), [machine-learning](), [new-releases](), [open-source](), [open-weights](), [security](), [software-engineering](), [staff](), [tech-news](), [technology](), [transformers](), [vllm](), [vulnerability](), [vulnerability-research](), [z-ai]() ### AI overview Cantina Security and Yeta Labs released apex-flash-1, an open-weights model fine-tuned for vulnerability research. On Cantina's internal held-out benchmark, it solved 40 of 60 tasks, compared with 36 for the GLM-5.3-Flash base model and 43 for Claude Opus 5 High. Its reported run cost was about $2.38, and BF16 deployment requires roughly 640 GB of GPU memory. ### Source excerpt Cantina Security, with Yeta Labs, has released apex-flash-1, an open-weights model trained specifically for vulnerability research. It is a reinforcement learning fine-tune of Z.ai's GLM-5.3-Flash, released on Hugging Face under the MIT license. Is it deployable? Yes, the MIT weights serve on vLLM, SGLang or Transformers, but BF16 needs roughly 640 GB of GPU memory. [...] The post Can an Open Model Do Security Research? Cantina's apex-flash-1 Solves 40 of 60 Held-Out Bug Tasks appeared first on MarkTechPost. ## Which AI models can you automate on Zapier? (OpenAI, Anthropic, Google, Moonshot AI, Z.ai, and more) DevFeed: [Which AI models can you automate on Zapier? (OpenAI, Anthropic, Google, Moonshot AI, Z.ai, and more)]() Original publisher: [Read original article]() Author: Nicole Replogle Published: 2026-09-23T05:00:00Z Content type: tutorial Language: en Sources: [Zapier]() Topics: [AI Models](), [Claude](), [AI Chat](), [open-responses]() Tags: [ai](), [ai-models](), [anthropic](), [google](), [mcp](), [moonshot](), [moonshot-ai](), [openai](), [z-ai](), [zapier-automation]() ### AI overview This guide lists AI models available through Zapier's built-in AI step and direct provider integrations. It describes how the options differ, including account and billing arrangements, and points readers to AutomationBench for comparing model performance on business workflows. ### Source excerpt Staying up to date on the latest AI models is like trying to keep my house vacuumed in the height of my dog's shedding season (a losing battle). New AI models drop practically every week, and it's not exactly obvious which ones deserve a spot in your workflows. This guide is your cheat sheet, with every model you can automate on Zapier right now, sorted by provider. If you just want to cut to the chase and find out which model you should be using for which task, that's what AutomationBench is fo ## China's Z.ai disables AI coding assistant features after security issue DevFeed: [China's Z.ai disables AI coding assistant features after security issue]() Original publisher: [Read original article]() Author: Reuters Published: 2026-09-22T03:23:25Z Content type: news Language: en Sources: [Technology | The Indian Express]() Topics: [coding assistant](), [Security](), [Vulnerabilities](), [Artificial Intelligence](), [data](), [Cloud](), [sensitive data](), [passwords](), [private key](), [Back end]() Tags: [ai](), [ai-coding-assistant](), [ai-coding-assistant-security-risks](), [ai-security](), [ai-security-risks](), [alibaba-cloud-data-privacy-controversy](), [artificial-intelligence](), [code](), [coding-assistant](), [encrypted](), [files](), [git-repository-cloud-upload](), [glm-5-3-ai-model](), [issue](), [it](), [nsfocus-independent-security-assessment](), [passwords](), [security](), [security-breach](), [technology](), [technology-artificial-intelligence](), [z-ai](), [z-ai-zcode-privacy-incident](), [z-ai-zero-data-retention-policy](), [zcode-codebase-indexing-vulnerability](), [zcode-unauthorized-code-upload](), [zhipu-ai-security-breach]() ### AI overview Z.ai disabled features in its ZCode AI coding assistant after reports that the default Codebase Indexing feature uploaded entire local code repositories to overseas cloud servers without users' consent. The company said it patched the vulnerability and plans to establish a product security vulnerability reporting and response process. ### Source excerpt The rare public disclosure of a security breach by a Chinese AI lab comes amid global warnings about frontier AI security risks. ## GLM-5.2 Cost Math: When Open-Weights Coding Models Actually Save You Money DevFeed: [GLM-5.2 Cost Math: When Open-Weights Coding Models Actually Save You Money]() Original publisher: [Read original article]() Author: Developers Digest Published: 2026-06-17T00:00:00Z Content type: article Language: en Sources: [Developers Digest]() Topics: [Artificial Intelligence]() Tags: [ai-coding-tools](), [ai-models](), [coding](), [cost](), [glm](), [glm-5-2](), [gpt](), [gpt-5-5](), [guide](), [math](), [models](), [open-weights](), [pricing](), [swe-bench](), [token-cost](), [z-ai]() ### AI overview An analysis of Z.ai's GLM-5.2, a 753B open-weights coding model, including its reported SWE-bench Pro comparison with GPT-5.5, per-token cost, cost-per-task example, and model-selection guidance. ### Source excerpt Z.ai's GLM-5.2 lands as a 753B open-weights coding model that beats GPT-5.5 on SWE-bench Pro for roughly one-sixth the per-token cost. Here is the real cost math, a worked cost-per-task example, and a when-to-use-which decision guide.