[CLSTR]
< Back to situations

Monitor this situation.

[SITUATION] · [QUIET] · [TECHNOLOGY]

9 clusters · 49 sources · 35 days · First seen · Last updated

Zhipu AI release of GLM-5.3 and GLM-5.3-Flash models

Overview

Zhipu AI has released GLM-5.3 and its optimized sibling, GLM-5.3-Flash, as open-weight models. GLM-5.3 utilizes a 744-billion-parameter Mixture of Experts architecture optimized for programming and cybersecurity, demonstrating state-of-the-art performance on benchmarks like Terminal Bench 3.0 and achieving an 84.5% success rate in CyberGym. While it excels at vulnerability identification, it has trailed Western competitors in complex exploit chain execution. The GLM-5.3-Flash variant features a 320-billion-parameter Mixture of Experts architecture with 18 billion active parameters per token. It is a multimodal model capable of natively processing text, images, and video, utilizing a hybrid architecture of linear and sparse attention to manage its 1-million-token context window. Before its official release, the model was distributed anonymously as ‘ox-alpha’ on platforms such as OpenRouter, where it became the most used model within a single week, reportedly exceeding 50 trillion tokens in traffic. On the Artificial Analysis Intelligence Index, GLM-5.3-Flash secured a top-ten position, performing similarly to Anthropic’s Claude Opus 4.8. A significant technical milestone is the model’s reliance on domestic Chinese compute infrastructure. Its online inference traffic is serviced by an array of approximately 100,000 domestically manufactured AI chips, such as those from Cambricon; analysts suggest the hardware may belong to Huawei’s Ascend series. Following the release, the US startup Abliteration.ai developed a modified version titled ‘abliterated-model-large-v2’. This ‘abliteration’ process modifies model weights to suppress internal activation patterns associated with refusal mechanisms, aiming to create a version less likely to decline sensitive or security-related queries while maintaining coding and cyber capabilities. Zhipu AI has since expanded the lineup with the launch of GLM-5.3-FlashX. This high-speed version is designed for enterprises and developers, offering a maximum inference speed of up to 200 tokens per second. While it maintains the 320-billion-parameter architecture and 1-million-token context window of the standard Flash model, it is priced higher.

Entities

Zhipu AI · GLM-5.3 · Z. ai · Anthropic · GLM-5. 3

Timeline

  1. [TECHNOLOGY] 6 sources
    Zhipu AI launches GLM-5.3-FlashX with 200 tokens/s speed

    Zhipu AI launched GLM-5.3-FlashX, an optimized large language model capable of speeds up to 200 tokens/s using 100,000 domestic chips.

  2. [TECHNOLOGY] 2 sources
    Zhipu AI operates new GLM-5.3-Flash model on 100,000 Chinese chips

    Zhipu AI's new GLM-5.3-Flash model was powered by over 100,000 Chinese chips. Meanwhile, US startup Abliteration.ai is offering modified versions of the model with removed safety filters.

  3. [TECHNOLOGY] 3 sources
    Zhipu AI launches GLM-5.3-Flash model powered by domestic chips

    Zhipu AI launched GLM-5.3-Flash, a 320B parameter multimodal model powered entirely by 100,000 domestic Chinese AI chips.

  4. [TECHNOLOGY] 2 sources
    AI security: Malicious Qwen model found on GitHub and Z.ai releases GLM-5.3

    Security researchers warned of a fake Qwen AI model on GitHub spreading the StealC virus, while Z.ai released its cyber-capable GLM-5.3 model with usage restrictions for major cloud providers.

  5. [TECHNOLOGY] 11 sources
    Z. ai releases GLM-5. 3 open-weight models

    Z. ai released the open-weight GLM-5. 3 and GLM-5. 3-Flash models, achieving performance comparable to Claude Opus 4.8 while running on domestic Chinese AI chips instead of Nvidia hardware.

  6. [TECHNOLOGY] 4 sources
    Zhipu AI releases GLM-5.3 with enhanced cybersecurity and coding capabilities

    Zhipu AI has launched GLM-5.3, a new model featuring significant advancements in cybersecurity reasoning and coding capabilities, now available for evaluation via ZenMux.

  7. [TECHNOLOGY] 4 sources
    Z. ai GLM-5. 3 demonstrates autonomous vulnerability research

    Z. ai’s GLM-5. 3 open-weight model has demonstrated autonomous offensive capabilities, including the ability to independently research and exploit vulnerabilities in the Cursor AI-native IDE.

  8. [TECHNOLOGY] 10 sources
    Zhipu AI releases GLM-5.3 to rival Anthropic in coding and security

    Chinese startup Zhipu AI released GLM-5.3, an open-weight AI model that rivals Anthropic in cybersecurity detection and coding, though it trails in functional exploit capabilities.

  9. [TECHNOLOGY] 15 sources
    Zhipu AI launches GLM-5. 3 with enhanced coding and cyber capabilities

    Zhipu AI launched GLM-5. 3, an open-weight model that uses enhanced post-training to significantly boost coding and cybersecurity performance, rivaling top US-based AI models.

Sources

arkade.com.br · blog.quintarelli.it · blogspan.net · classics.itmedia.co.jp · cnmo.com · cryptobriefing.com · cryptopolitan.com · detlionblood32.wordpress.com · dev.to · digitalphablet.com · drweb.de · elektronikpraxis.de · exame.com · flagthis.com · gigazine.net · hackernews.com · hothardware.com · ihal.it · infoq.cn · inversion.es · ipaddisti.it · ithome.com · kienthuc.net.vn · leiphone.com · m.scmp.com · mayacomunicacion.com.mx · memeburn.com · news.az · news.yesky.com · newsbytesapp.com · newscenter.io · noticiasdemalaga.es · pressadvantage.com · prtimes.jp · slashgear.jp · solidsoftwaretools.com · spacemoney.com.br · sportguide.ch · tech360.tv · tecnoandroid.it · the-decoder.de · thenewstack.io · theregister.co.uk · tmtpost.com · together.ai · tokenpost.kr · tomshw.it · unite.ai

This summary has been updated 12 times: see revision history