GLM

Evidence (20)

• 7Simon Willison·9/29/2026research
  Anthropic red team cites advanced exploit capability in GLM-5.3

  A quotation from Anthropic's Frontier Red Team reports that GLM-5.3 achieved full control-flow hijacks in 4% of trials on an internal binary exploitation benchmark. The quote says this crosses a meaningful threshold, since earlier models such as Claude Opus 4.6 and GLM-5.2 did not succeed on any trials.

• 8Hacker News·9/29/2026security
  GLM-5.3 Shows Advanced Cyber Exploit Capability

  A Frontier Red Team Policy post evaluates GLM-5.3, an open-weight model from Zhipu AI, and finds it can autonomously build end-to-end cyber exploits at a level comparable to Claude Mythos Preview. The report also says GLM-5.3’s safeguards are weak and can often be bypassed with simple techniques, increasing the risk that malicious actors could use it for advanced attacks.

• 7Hacker News·9/23/2026research
  AI Token Costs Are Falling Fast

  The article argues that the cost of using machine learning, especially large language models, is dropping by orders of magnitude and may keep falling. It cites improvements in GPUs, model efficiency, hosted inference, local models, and specialized use cases, and predicts AI will become infrastructure embedded across computing.

• 9Hacker News·9/18/2026security
  Report says ZCode silently uploads Git workspaces

  A reverse-engineering report claims the ZCode AI coding desktop app uploads users’ full workspaces, including .git history and configs, to Aliyun OSS whenever it is logged in. The report says the archive is encrypted with a server-wrapped key that the local client cannot decrypt, and that UI settings do not fully stop the collection and upload flow.

• 7The New Stack·9/15/2026product_launch
  Bolt offers 50x more model usage in exchange for training data

  Bolt.new is testing Forge, a research preview for Pro subscribers that provides up to 50 times more usage of open-weight coding models through October 14. In return, users must opt in to sharing anonymized coding sessions, including prompts, source code, fix traces, and agent conversations, which Bolt and Arcee AI will use to train a future open-weight model.

• 7The New Stack·9/14/2026product_launch
  OpenRouter adds US-only routing for business AI traffic

  OpenRouter has launched general availability of US in-region routing for business and enterprise users, ensuring requests sent through its US endpoint are decrypted, processed, and served entirely inside the United States or rejected. The move responds to growing use of open-weight models, especially Chinese models, in US token traffic and gives companies more control over data residency.

• 8Hacker News·9/8/2026security
  Open-weight AI models raise urgent security concerns

  The post argues that newly available open-weight models such as GLM 5.3-flash can be run locally and adapted to remove safety refusals, making powerful offensive capabilities more accessible. It calls for rapid, industry-wide vulnerability fixing using frontier LLMs, before these models are widely used for malicious hacking. The piece also emphasizes that deployment of fixes, not just finding them, is the hardest part.

• 4Hacker News·9/8/2026research
  Benchmarking model and harness combos on a Three.js task

  A Hacker News post reports tests of 10 model and harness combinations on the same prompt to build a single-page Three.js sci-fi hangar scene. The write-up compares runtime, token usage, tool calls, browser checks, and estimated costs across models such as GLM 5.3 Flash Max, GPT-5.6 variants, and Qwen 3.8 27B under different harnesses.

• 5The New Stack·9/1/2026model_release
  GLM-5.3-Flash Trades Time for Lower Cost

  The New Stack compared Z.AI’s GLM-5.3-Flash and GLM-5.3 on coding, reasoning, and information-extraction tasks. Both models produced correct results, but Flash used far more tokens and time while still costing less per request due to much lower pricing.

• 8Hacker News·8/28/2026open_source
  GLM-5.3 released as open weights

  GLM-5.3 is now available as an open-weights model, built on the same base model as GLM-5.2 with improvements coming entirely from post-training. The release highlights stronger coding and long-horizon task performance, along with notable gains in cyber-related benchmarks and local deployment support across several inference frameworks.

• 6Hacker News·8/27/2026research
  Small, Fast AI Models Reach the Pareto Frontier

  The article argues that newer small models such as gpt-5.6-luna and GLM 5.3 are now fast, capable, and cheap enough to be useful for many everyday and business tasks. It suggests these models could unlock more consumer AI products and expand adoption for responsive, good-enough automation in companies.

• 7Hacker News·8/26/2026model_release
  GLM-5.3-Flash

  GLM-5.3-Flash is a newly discussed model release on Hacker News, drawing strong attention from the developer and AI community. The discussion suggests interest in its capabilities, performance, and practical use compared with other current models.

• 7Hacker News·8/26/2026model_release
  Z.ai confirms Ox Alpha as a new GLM model and plans to open its weights

  Z.ai said Ox Alpha is a new model in its GLM family and that it will release the model weights. The announcement suggests the company is positioning the model as a competitor in the open-weight large language model space.

• 7The New Stack·8/24/2026model_release
  Ox Alpha Raises Privacy and Attribution Questions

  Ox Alpha has appeared as an anonymous coding model on OpenRouter and OpenCode, with strong speculation about who built it but no confirmed attribution yet. The bigger issue is that different access routes appear to carry different data-retention and training terms, creating uncertainty for developers sending private code through the model.

• 3Simon Willison·8/23/2026research
  Drew Breunig on Fable and the end of the free lunch

  Simon Willison quotes Drew Breunig arguing that after Fable arrived, it no longer made sense to ignore coding harnesses or context strategies and simply wait for better models. The note suggests that model quality and pricing no longer automatically cover shortcomings in workflow design, so teams must think more carefully about where work is done.

• 7Hacker News·8/23/2026breakthrough
  GLM-5.3 tops real-world LLM benchmark at lower cost

  A Hacker News post highlights a benchmark comparing LLMs on 28 practical tasks, with GLM-5.3 reportedly becoming the first model to clear all five categories at 100%. The write-up says it outperformed Anthropic and OpenAI models on this suite while costing about one-fifth as much, though it had a slower median time-to-first-token than some alternatives.

• 4Hacker News·8/23/2026security
  AI models help root an Amazon Fire HD 10 tablet

  A user reported spending $266 on AI model usage to help find and validate a root exploit for an Amazon Fire HD 10 tablet used as a kiosk. The effort centered on a known Mali GPU kernel vulnerability, CVE-2022-38181, which had been patched by Amazon in later firmware but remained exploitable on the user's version. The write-up describes how multiple AI models assisted with analysis after conventional methods and months of manual investigation had stalled.

• 5Hacker News·8/18/2026research
  GLM-5.3 Scores Strong on Artificial Analysis Benchmarks

  Artificial Analysis published benchmark results for GLM-5.3 (max), a proprietary reasoning model with a 1M-token context window. It ranks highly on intelligence, runs faster than average, and is priced in the mid-range for comparable models, though it is notably verbose.

• 7The New Stack·8/18/2026security
  OpenAI Warns Z.ai’s GLM-5.3 Could Worsen Cyber Risk

  OpenAI president Greg Brockman said Z.ai’s forthcoming open-weight GLM-5.3 model could significantly accelerate the cyber threat landscape. He contrasted Z.ai’s planned public weight release with OpenAI’s more restricted access to its own GPT-5.6-Cyber model, which is limited to vetted users under its cyber safety program.

• 8The New Stack·8/14/2026model_release
  GLM-5.3 claims frontier coding gains and cyber capabilities

  Z.ai has released GLM-5.3, presenting it as a frontier coding model with emergent cyber capabilities. The announcement has drawn substantial attention on Hacker News, suggesting strong interest in both its coding performance and potential security implications.