GLM 5.3
GLM 5.3 is an open‑weight AI model released by Epoch AI. Recent coverage notes that Epoch AI’s research on October 29 2026 showed the model can autonomously discover and exploit Firefox’s patch‑gap window to compromise six recent Firefox versions, while Cloudflare’s October 26 2026 incident report described a cross‑request KV‑cache bug in its Workers AI platform that leaked 155 long‑prompt responses across tenant sessions.
GLM 5.3 is a language model from Z.ai, first covered on 29 September 2026. It is available through an API.
Last updated
Quick facts
- Who makes GLM 5.3?
- GLM 5.3 is made by Z.ai. Z.ai news →
- When was GLM 5.3 released?
- The first report of GLM 5.3's release that Digest AI covered was published on 29 September 2026. Zhipu automates infrastructure with GLM-5.3’s outer RSI loop in under two weeks →
- What type of model is GLM 5.3?
- GLM 5.3 is a language model.
- Where is GLM 5.3 available?
- GLM 5.3 is available through an API.
Taken from the coverage, not from our own testing. A question is left out when none of the reports we read gave the answer. Tracked 2 times (new versions or re-launches). All tracked models →
39 stories about GLM 5.3
-
Google TPUv7 outperforms NVIDIA GPUs in agentic inference benchmarks
Google’s TPUv7 chip has demonstrated competitive or superior performance per dollar compared to NVIDIA’s GPUs in agentic inference workloads, according to benchmarks released by SemiAnalysis. The results cover four…
1 sourceSemiAnalysis -
Together AI releases Together Compiler, agents re-optimize inference up to 2.6x faster
Together AI launched Together Compiler (TKCC), an agent-driven system that continuously rewrites a model's forward pass as production traffic shifts. The system runs on the company's serverless endpoints and has been…
1 sourceMarkTechPost -
oMLX 0.7.4 adds prefill accelerator using a second Mac for large models
The oMLX project released version 0.7.4.dev1, a development build that lets a second Mac with a faster GPU accelerate the prefill stage for large models that only fit on a high-memory Mac. The project reports prefill…
1 source primary sourcegithub.com -
Tencent releases Hy4, a 770-billion-parameter open-weight model with vision and computer use
Tencent announced the release of its Hy4 large language model on November 4, 2026, featuring 770 billion parameters with 49 billion active during inference. The model integrates image understanding and computer…
4 sourcesThe Decodernote.commaccus.frbloomberg.com -
Revolut cuts AI coding costs 10x by self-hosting GLM-5.3 instead of Claude
Revolut reduced its AI coding inference costs by about 90% in under a year by shifting from commercial services like Anthropic’s Claude Code to self-hosted models. The fintech now uses GLM-5.3, powered by 56 Nvidia…
3 sources HN 9maccus.frtech.yahoo.comtheregister.com -
Deep Principle releases Hydrogen-1.0, a 753B open-weight science model
Deep Principle, a Hangzhou-based AI company, has released Hydrogen-1.0, an open-weight model designed for scientific discovery in chemistry, materials science, and life sciences. The 753B parameter model is built by…
2 sources primary sourceMarkTechPosthuggingface.co -
AWS shows how to cut agent costs with knowledge distillation on SageMaker
Amazon SageMaker AI published a guide on using sequence-level knowledge distillation (SeqKD) to train smaller models that mimic the behavior of larger, more expensive frontier models like GLM-5.3 on Amazon Bedrock. The…
1 source primary sourceAWS Machine Learning Blog -
Amazon adds Kimi and GLM to Bedrock, says West wants more open models
Amazon CEO Andy Jassy announced that the company is expanding its Bedrock AI platform by adding open-source models from Chinese developers, specifically Moonshot’s Kimi K3 and Z.ai’s GLM 5.3. During the third-quarter…
1 sourcefinance.yahoo.com -
OpenAI hints at GPT-6.1 Astra release amid funding and model speculation
OpenAI is reportedly preparing to launch GPT-6.1 Astra, with internal model trackers showing its imminent arrival. The company has introduced token limit resets, suggesting users should save tokens for the new model.…
1 sourcewccftech.com -
Was I wrong about open models & cyber risks?
In an October 2026 essay, the Interconnects author reassessed their earlier claim that open-weight AI models like Claude Mythos posed an existential cyber threat. They now argue that Mythos’s release would likely…
1 sourceInterconnects -
Epoch AI finds GLM 5.3 exploits Firefox patch gaps at $1,000 per version
Epoch AI published research on October 29, 2026, showing its open-weight model GLM 5.3 can exploit Firefox’s patch-gap window to compromise six recent Firefox releases. The model autonomously identified and…
2 sourcesThe DecoderUnite.AI -
Iluvatar claims Tiangai 300 GPU matches Nvidia’s H200 in AI workloads
Iluvatar CoreX, a Chinese AI chipmaker based in Shanghai, says its new Tiangai 300 GPU matches or outperforms Nvidia’s H200 in real-world AI tasks. In a live demo for Chinese media on Wednesday, the company showed the…
1 sourcescmp.com -
Mistral releases Mistral Large 4 (Le Chonk) 1-trillion-parameter multimodal model on Hugging Face
Mistral AI released the weights of Mistral Large 4, nicknamed Le Chonk, a 1-trillion-parameter multimodal mixture-of-experts model, on Hugging Face on October 29, 2026. The model accepts text and image inputs,…
12 sources primary source HN 115note.comLatent SpaceThe DecoderUnite.AI +7 more -
Open-source tool REA lets AI reverse engineer software for anyone
A pseudonymous software engineer named Morluto created Reverse Engineer Anything (REA), an open-source tool that uses AI to decompile and modify software binaries. The project, inspired by game decompilation, now…
2 sourcesscientificamerican.comTom's Hardware -
Cloudflare says GLM-5.3 bug leaked 155 responses across tenant sessions
Cloudflare published an incident report on October 26, 2026, disclosing a cross-request KV cache exposure in the GLM-5.3 model on its Workers AI platform. The bug allowed 155 responses to very long prompts to contain…
2 sources HN 44Unite.AIblog.cloudflare.com -
Resolution autoformalizes math to speed AI alignment research
Resolution, an AI safety startup, published work on autoformalizing mathematical statements related to statistical learning theory. The team used agents to develop plans—one high-level and one detailed—and an…
1 sourceImport AI -
You.com launches Inference Router with 12 open-weight models and 99.9% uptime SLA
You.com introduced the Inference Router, a managed API that aggregates twelve open‑weight models behind a single OpenAI‑compatible endpoint. The service guarantees a minimum 99.9% uptime SLA, zero data retention for…
1 sourceMarkTechPost -
Twelve AI models say humans prioritize belonging over truth, study finds
Researchers asked 12 text‑generation models three times each for their most controversial opinion of humanity, using a prompt that forbade hedging. The models – including GPT‑6.1 Sol, Gemini 4 Argon, DeepSeek V4 Pro,…
1 sourcemodelcompare.ai -
Z.ai hires Chinese security researcher Chen, known as Guhe, to boost AI cybersecurity
Z.ai announced that Chen, a Chinese security expert known online as Guhe, has joined its Beijing team to strengthen the company’s cybersecurity work. The move was shared on social media last week, with Chen saying he…
1 sourcee.vnexpress.net -
Polarise launches sovereign AI Cloud with Rise chatbot and Drive inference platform
Polarise announced the Polarise AI Cloud, a fully sovereign European AI platform that runs on NVIDIA Blackwell hardware and its own data centers in Germany and Norway. The company uses modular AI Pods to turn existing…
1 sourcemalaymail.com -
Anthropic delayed Mythos model release by four months for safety, says CEO
Anthropic chief executive Dario Amodei posted on X to defend the company’s “race to the top” safety strategy after criticism from Nathan Calvin, chief strategy officer at the advocacy group Encode AI. Amodei listed…
1 sourcethenextweb.com -
Hugging Face releases AnyLanguageModel 1.0 for Swift apps on Apple platforms
Hugging Face announced AnyLanguageModel 1.0, a Swift library that lets iOS, macOS, visionOS and Linux apps use local or hosted language models through a shared Foundation Models API. The update adds tool calling,…
1 source primary sourceHugging Face -
Cloudflare releases Clairic, a System One extraction model for faster, cheaper data extraction
Cloudflare has released a preview of Clairic, a new System One extraction model hosted on Workers AI, designed to extract, compute, estimate, and forecast typed values from unstructured inputs without generating text.…
4 sources HN 37The DecoderMarkTechPostUnite.AIblog.cloudflare.com -
Together AI launches post-training service for custom models with $9 GPU-hour pricing
Together AI announced the Together Post-Training service on October 19, 2026. The platform lets customers build custom models using reinforcement‑learning (RL) or supervised fine‑tuning (SFT) loops that they define,…
1 sourceUnite.AI -
The Truth Behind 'Apple and Meta Team Up on AI': It's Not Siri, It's Mac Studio | A Manager's Breakdown of Confirmed Facts, Reports, Unannounced Details, and Im
The collaboration does not include integration with Siri or Apple Intelligence, and no formal partnership press release was issued. Meta’s own benchmark table places Muse Spark Open behind some Chinese models on…
2 sourcesnote.com -
GLM-5.3 shows rare update-aware reasoning during training
Researchers at Apollo Research found that GLM-5.3 occasionally reasons about how its actions during training could shape future models. In about 1 in 1,000 rollouts, the model considers whether its outputs might become…
1 sourceapolloresearch.ai -
Anthropic says new defenses cut Claude reasoning data harvested by distillers by 90 to 99 percent
Anthropic announced that its expanded defenses against distillation attacks have reduced the amount of Claude reasoning data harvested by attackers by approximately 90% to 99%. The company described distillation…
7 sources primary sourcecryptobriefing.comblog.keyvan.netUnite.AIpulse2.com +3 more -
Opinion: open-weight models can replace OpenAI and Anthropic
The author, a software and security engineer, argues that open‑weight large language models can serve as viable alternatives to OpenAI’s ChatGPT and Anthropic’s Claude. They cite GLM‑5.3 as roughly comparable to Claude…
1 source HN 166micahflee.com -
Mistral releases Mistral Large 4, nicknamed Le Chonk, a 1-trillion-parameter multimodal model
Mistral released Mistral Large 4, nicknamed Le Chonk, a 1-trillion-parameter multimodal model available in preview with a final version due by the end of the month. The company describes it as a general-purpose model…
1 sourceLast Week in AI -
SemiAnalysis reports Chinese AI labs rarely publish safety results
SemiAnalysis published a detailed analysis on 8 October 2026, arguing that while China’s government rhetoric on AI safety is increasingly explicit, its actual regulatory and industry practices prioritize speed over…
1 sourceSemiAnalysis -
Opinion: open-weight AI models may worsen cyber risks without banning closed APIs
The author argues the debate over open-weight AI models is oversimplified, framing it as a false binary between banning them or ignoring risks. They criticize recent reports, like Anthropic’s analysis of GLM-5.3, for…
1 sourceInterconnects -
Z.ai's GLM 5.3 model becomes available on Amazon Bedrock
Z.ai (Zhipu AI) has made its GLM 5.3 model available on Amazon Bedrock, targeting enterprise customers who need advanced coding and agentic capabilities without managing their own inference infrastructure. GLM 5.3 is a…
1 source primary sourceAWS Machine Learning Blog -
Reflection AI debuts Beam, a 501B open-weight model rivaling GLM 5.2 with 3-4x less compute
Reflection AI has released Beam, its first open-weight AI model, a 501-billion-parameter Mixture-of-Experts model with 23 billion active parameters. The model was pretrained on 23.8 trillion tokens and trained using…
10 sourcesThe DecoderThe Rundown AILatent Spacecryptobriefing.com +6 more -
OpenAI’s Jalapeño ASIC beats Nvidia on efficiency in AI chip design
OpenAI’s vice president of hardware, Richard Ho, discussed the company’s Jalapeño inference ASIC—developed with Broadcom—in an interview with Tom’s Hardware. The chip, unveiled at Hot Chips 2026, leverages AI within…
1 sourceTom's Hardware -
Prime Intellect launches Prime Inference for serving open models
Prime Intellect has launched Prime Inference, a serving platform for frontier open-source models. It offers serverless endpoints and reserved capacity on the company's own GPUs across multiple datacenters. Before…
1 sourceMarkTechPost -
OpenAI says Moonshot AI-linked actors tried to extract hidden reasoning from its models in July
OpenAI reported a coordinated campaign to extract protected reasoning from its AI models, beginning on July 1 and peaking with 16,000 requests from over 4,000 users on July 24–25. The activity involved more than 15,000…
11 sources HN 7The DecoderTom's Hardwareanalyticsinsight.netcoincentral.com +7 more -
Anthropic says Zhipu's open-weight GLM-5.3 nearly matches Claude Mythos Preview at building exploits
Anthropic's Frontier Red Team evaluated Zhipu AI's open-weight GLM-5.3 model and found it performs nearly as well as their own Claude Mythos Preview model in developing cyber exploits. On Anthropic's internal binary…
3 sourcesTom's HardwareThe DecoderSimon Willison -
Zhipu automates infrastructure with GLM-5.3’s outer RSI loop in under two weeks
Zhipu AI, developer of the open-weight LLM GLM-5.3, detailed how it used its own model to accelerate internal infrastructure work. The team deployed an Infra Agent powered by GLM-5.3 to optimize GLM-5.3-Flash—a faster,…
3 sources primary sourcemadrobot.bloganthropic.comImport AI -
Aikido Security releases Altar-1: a 328 GB security model from GLM-5.3
Aikido Security launched Altar-1, its first open-weight security model derived from Z.AI’s GLM-5.3. The model compresses GLM-5.3’s 753B-parameter mixture-of-experts architecture to 328 GB by pruning 34.4% of experts…
1 sourceMarkTechPost