TokenPost topic
GLM-5.3
7 TokenPost stories are tagged GLM-5.3. The newest appear first.
Latest GLM-5.3 stories
- NVIDIA Runs 753-Billion-Parameter GLM-5.3 Across Two DGX Stations
The vLLM setup reached 607.8 tokens per second with 16 requests running at once on the SPEED-Bench Coding dataset.
· Technology
- Revolut Says GLM Tuning Lifted Inference Speed 88% on 56 B200 GPUs
Abitbol also cited GLM-5.1 at about $0.30 per million input or output tokens.
· Technology
- Revolut Makes GLM 5.3 Its Default AI Model for About 2,000 Engineers
The setup uses about 56 Nvidia B200 GPUs at a blended cost of about $0.30 per 1 million tokens. A coding-time metric improved about 60%, while a task-delivery measure improved about 70%.
· Technology
- Hydrogen-1.0 Tops GPT-6 Astra With a 38.3% Research Pass Rate
Built on GLM-5.3, the model was trained on laboratory records and demonstrated simulated chemistry and materials-screening tasks.
· Technology
- GLM-5.3 Completed Three of 13 Simulated Cyberattack Scenarios
The exercises recorded no refusals on offensive tasks and included automated scripting and data extraction.
· Technology
- GLM-5.3 Nearly Matches Anthropic Model in Cyber Exploit Tests
The open-weight model produced working exploits in 50 of 410 attempts, while U.S. testing ranked it the most capable open-weight model evaluated to date.
· Technology
- Anthropic Says GLM-5.3 Can Build End-to-End Network Exploits
Simple simulated attacks bypassed the model’s safeguards about 64% to 100% of the time, while protected Claude models resisted the same methods.
· Technology