Qoder ships GLM-5.2 — open-source finally cracks long-horizon agentic tasks

Qoder now supports GLM-5.2, the latest open-source flagship model.

What GLM-5.2 brings:

  • Solid 1M context that holds up in practice
  • Open-source SOTA on Code Arena overall benchmark
  • Purpose-built for Long Horizon Task reliability in Agentic Engineering

As AI Agents scale, models need to sustain coherence across thousands of tool calls, tens of thousands of lines of code, and massive intermediate state. Open-source models have matched closed-source on short tasks — but long-horizon execution is where they consistently broke down. GLM-5.2 is a meaningful step forward here.

To enable: Model selector → GLM-5.2strong text