GLM 5.1 advances Z.ai's GLM-5 generation with a focus on long-horizon autonomous coding. It can work independently on a single task for over eight hours, planning, executing, and iterating until it delivers engineering-grade results.
import { streamText } from 'ai'
const result = streamText({ model: 'zai/glm-5.1', prompt: 'Why is the sky blue?'})GLM 5.1 builds on the GLM-5 generation with a significant jump in coding capability, released April 7, 2026. Where GLM-5 introduced multiple thinking modes and agentic workflows, GLM 5.1 pushes the autonomy envelope: it sustains focus on one task for over eight hours, continuously planning, writing code, running tests, and improving its own output without human intervention.
The model targets long-horizon tasks that earlier models struggle with. Multi-file refactors, end-to-end feature implementation, and large-scale codebase migrations benefit from the extended autonomous execution window. Rather than handing back partial results for human review at each step, GLM 5.1 completes the full loop and delivers finished, tested code.
GLM 5.1 supports a context window of 204.8K tokens and max output of 202.8K tokens. Through AI Gateway, it shares the same unified API, built-in observability, and provider routing as other Z.ai models.