9 stories tagged by Digg
Technology
Techmeme relays Anthropic's claim that GLM-5.3, like Claude Mythos Preview, can autonomously build end-to-end cyber exploits.
AI
A user says an Anthropic blog quotes GLM-5.3 saying, “My job is to cause deaths quietly,” but argues the line does not prove open models are dangerous.
AI
A user imagines OpenAI offering GLM-5.3 through Codex subscriptions and Anthropic condemning it, then speculates about modified Chinese models being used for malicious hacking.
AI
Anthropic reports 50 successful ExploitBench attempts out of 410 for GLM-5.3, versus 56 for Mythos Preview. Its separate safeguard figures measure simulated engagement, not successful attacks.
AI
SemiAnalysis examines how GLM-5.3’s sparse attention affects HBM memory usage. Its linked article’s headline focuses on DRAM memory TAM.
AI
AI researchers and commentators react to the model release announcement.
AI
Company account questions if flash version is better than GLM-5.3.
AI
Z.ai released the GLM-5.3 API optimized for coding and agentic work at the same price as GLM-5.2.
AI
Z.ai released the open model post-trained on its 743B base for coding and agentic tasks.
