Don't fall behind, keep up with latest AI news here
Hacker News
Read full postGLM-5.3 matches top open weights performance with a 246-point agentic work gain but uses 20% more output tokens per task. OpenAI halved Sol pricing on OpenRouter, Anthropic extended Claude Code limits, and IBM found agent memory benefits vary by model.



