On 16 June 2026, Z.ai announced GLM-5.2. Z.ai made sustained long-horizon work the centre of GLM-5.2 and published unusually detailed agent-evaluation settings.
This release brief was checked against first-party material on 10 August 2026. The date above is the public announcement date, not the date a repository was created or a third-party provider added the model. Where access or weights arrived later, that distinction is recorded below.
Release record
| Field | Verified detail |
|---|---|
| Announcement | 16 June 2026 |
| Availability or weight release | 16 June 2026 as MIT-licensed BF16 and FP8 checkpoints plus API access |
| Release type | open-weight frontier agent model |
| Access | MIT-licensed weights with no stated regional restriction, plus Z.ai services |
| Architecture | a sparse MoE using IndexShare attention indexing and improved multi-token prediction |
| Maximum stated context | one million tokens |
What changed
GLM-5.2 combines strong headline scores with an architectural efficiency claim. IndexShare reuses one indexer across four sparse-attention layers, which Z.ai says reduces per-token FLOPs by 2.9 times at one million tokens, while the revised draft layer increases speculative-decoding acceptance by up to 20%. The model card also documents timeouts, context sizes and harnesses for many agent tasks, making its numbers more useful than an unexplained leaderboard collage.
The practical comparison is therefore not simply whether GLM-5.2 has the largest headline score. Teams need to ask whether its architecture, access terms, latency, tool behaviour and evaluation setup match the workload they actually intend to run. A model can lead one harness while losing on cost, refusal behaviour, multilingual quality or repeatability in another.
Benchmarks worth retaining
| Evaluation | Reported result | How to read it |
|---|---|---|
| AIME 2026 | 99.2 | Reasoning result under the disclosed sampling setup |
| SWE-Bench Pro | 62.1 | OpenHands run with a tailored prompt and 400K context |
| Terminal-Bench 2.1 | 81.0 controlled; 82.7 best reported harness | Harness distinction preserved by the model card |
These are release-time results, not independently reproduced guarantees. Z.ai reports both a controlled Terminus-2 score and a best-harness score for Terminal-Bench; those values should not be substituted for each other. Scores should remain attached to the disclosed effort setting, agent harness, tool access, timeout, context-management policy and judge model. Moving a number into a procurement sheet without those conditions creates false comparability.
Architecture and access
GLM-5.2 is described as a sparse MoE using IndexShare attention indexing and improved multi-token prediction with one million tokens of stated context. Its access position at verification time is MIT-licensed weights with no stated regional restriction, plus Z.ai services. That wording matters: open weights, source-available weights, an API, a product preview and a research demonstration give adopters very different rights and different levels of reproducibility.
Before deployment, record the exact model identifier or checkpoint, inference stack, quantisation, reasoning setting, region, price schedule and supplier terms. If the release uses a custom licence, read the licence itself rather than relying on the word “open” in launch copy. If it is API-only, preserve the dated documentation and change-notice route because the served snapshot can change without a downloadable artefact.
What an evaluation should test next
For GLM-5.2, a credible internal gate should include:
- a frozen set of representative tasks with pass, fail and abstain criteria;
- a matched baseline using the same tools, timeout, prompt budget and reviewer rubric;
- repeated runs to expose variance rather than reporting a single best attempt;
- latency, token use and total task cost alongside task success;
- adversarial, multilingual and long-context cases relevant to the real deployment; and
- rollback evidence showing the previous model can be restored safely.
The wider model change-control guide explains how to keep model, prompt, tool and corpus changes reconstructable. The AI dependency inventory guide covers the release and supplier records needed after deployment.
AIEngine verdict
GLM-5.2 is one of the window’s strongest and best-documented open agent releases. The MIT licence and reproducible settings make it a priority evaluation candidate, provided infrastructure can support its scale.
This is a launch assessment, not a certification. Benchmark leadership is useful evidence of where to test; it is not authorization to place the model in a high-impact workflow without domain evaluation, security review and an accountable owner.
Primary sources
Image provenance
Hero image: Z.ai official model card via Hugging Face. The locally served WebP is a crop of the first-party release or model-card asset recorded in the repository provenance manifest.
