Board/Z.ai/GLM-5.2
Z

GLM-5.2

Zhipu AI (Z.ai)·Jun 13, 2026N/A

Open weights

MIT-licensed long-horizon specialist: lossless 1M-token context and the strongest open-weights SWE-bench Pro run of early summer (62.1%), built specifically for sustained multi-hour engineering tasks.

Status
Generally available
License
Open weights · MIT
Context window
1M
Max output
128K
Input · $/M
$1.40
Output · $/M
$4.40
Cached input · $/M
$0.14
Blended · $/M
$2.15
Parameters
744B MoE · ~40B active
Inputs
Text + images

Benchmarks

Composite 62.1· mean of published scores

Reasoning & knowledge

  • GPQA Diamond

    no public score

  • MMLU-Pro

    no public score

  • Humanity's Last Exam

    no public score

  • ARC-AGI-2

    no public score

Coding

  • SWE-bench Verified

    no public score

  • SWE-bench Pro

    62.1%

    dataset best 80
  • LiveCodeBench

    no public score

Agents & computer use

  • Terminal-Bench 2

    no public score

  • OSWorld 2.0

    no public score

  • BrowseComp

    no public score

Math

  • AIME 2026

    no public score

Known strengths

  • Best open SW-Pro at launch
  • Lossless 1M context
  • MIT license

Primary sources

  • Z.ai blog
Compare this model →

More from Z.ai

  • ZGLM-5.3Aug 14, 2026 · 1M ctx
  • ZGLM-5Feb 11, 2026 · 200K ctx