The GLM 5.3 API has introduced a breaking change by removing the ability to explicitly disable the "thinking" mode, which was available in GLM-5.2. Developers must now use "thinking": {"type": "enabled", "effort": "low"} as a substitute, though this may result in slightly higher latency and token usage compared to the previous disabled state. Despite this change, GLM 5.3 offers notable improvements in agentic and coding-specific tasks, with Zhipu AI reporting significant gains on benchmarks like Terminal-Bench 3.0. AI
IMPACT Developers must update their code to accommodate GLM 5.3's API changes, potentially impacting latency-sensitive applications.
RANK_REASON The item details a breaking change in an API for a specific model version, requiring code updates for users.
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →