White-hat hacker Yuki Chen, who topped Microsoft's MSRC security researcher rankings five times, announced on X that he has joined Zhipu, saying he hopes to leverage cutting-edge AI to help build a more secure cyberspace and mentioning Z.ai. He previously served as director of 360's Vulnerability Research Institute and CTO of CyberKunlun.
Zhipu has begun a grayscale test of its high-speed flagship model GLM-5.3 HighSpeed in its AI coding tool ZCode, randomly opened to ZCode subscribers without a formal application channel. The model has a 1 million token context window and can output up to 131,072 tokens at a time, with reported speeds exceeding 490 tokens/s, far higher than DeepSeek V4.1 Flash's 121.34 tokens/s and GLM-5.3's 70.84 tokens/s.
At a speech in Singapore, Zhipu founder Tang Jie outlined the R&D direction for the next-generation large models GLM-5.4/5.5, including expanding pretraining data and scale to several T, raising parameters from over 740 billion to at least 1 trillion, strengthening post-training methods like SFT, RLHF, and RLVR, and improving long-horizon reasoning and agent abilities through two new paths: RSI autonomous iteration and longer, more autonomous agents. He had previously defined the next-generation model as fully self-training (RSI) at an earnings call in late August.
Zhipu AI's GLM-5.3 model has been officially launched on Amazon Bedrock, providing enterprise users with coding and agentic capabilities, with access guides available in the official documentation.
Goldman Sachs raised its rating on Zhipu from Neutral to Buy, with a 12-month DCF-based target price of HK$1,560. It lifted its 2026 year-end ARR forecast to $3.2 billion and raised 2026-2028 revenue estimates by 3%-12%, with base/bull/bear valuations at $98B/$157B/$32B.
Anthropic released a report on September 29 stating that Zhipu's GLM-5.3 demonstrates strong cybersecurity capabilities, nearly matching Claude Mythos Preview in end-to-end exploit tests. The report also noted that GLM-5.3's guardrails could be bypassed in 64% to 100% of cases.
On September 28, Zhipu announced a compensation plan for the ZCode data upload controversy: paying users and those returning within one month will receive four weekly quota reset cards and four 5-hour quota reset cards, valid for one month; from September 28 to October 7, 100,000 grants of 100 million tokens will be distributed daily to all users. Zhipu reiterated that the repository snapshot upload pipeline has been removed and the cloud data involved has been deleted, as confirmed by third-party checks.
On September 22, the latest daily rankings from OpenRouter showed that Zhipu AI's GLM-5.3-Flash recorded 4.36 trillion tokens in daily call volume, a 73% increase from the previous period, ranking first on the platform.
Zhipu AI's Zixuan Li announced that a faster GLM-5.3-Flash is now live, reaching up to 200 tokens/s. The model code is glm-5.3-flashx, priced at 2.5x the standard GLM-5.3-Flash on both the Coding Plan and API, and is open to all API users. Coding Plan users can apply via a form.
On September 21, Zhipu announced that its AI coding tool ZCode will be open-sourced soon, as a first step in addressing external data privacy concerns. Zhipu has invited CAICT and NSFOCUS to conduct security audits, confirming that the zcode-prod Alibaba Cloud OSS bucket contains zero cloud data and that all data objects and the bucket have been deleted. The ZCode v3.14.0 client has completed remediation, removing the Repo Wiki feature and cutting off local repository snapshot upload links. Zhipu also said it will establish a routine security vulnerability reporting mechanism.