GLM-5.2
Z.ai · Released 16 Jun 2026 · glm-5.2
$1.40
Input price / 1M tokens
#32 of 61, cheapest first
$4.40
Output price / 1M tokens
#27 of 61, cheapest first
1.05M
Context window
#11 of 71
88
Output tokens / second
#34 of 61
62.1%
SWE-bench Pro
#6 of 16
81%
Terminal-Bench 2.1
#11 of 17
91.2%
GPQA Diamond
#13 of 27
40.5%
Humanity’s Last Exam (no tools)
#10 of 23
Summary
GLM-5.2 is an open-weight model from Z.ai, released on 16 Jun 2026.
At $1.40 input and $4.40 output per million tokens, it is #32 of 61 on input price, cheapest first.
Its 1.05M-token context window ranks #11 of 71.
Measured output speed is 88 tokens per second, #34 of 61.
Z.ai flagship for long-horizon tasks with a 1M-token context, 744B total and 40B active parameters. Open weights under plain MIT with no regional limits (commercial use allowed). Superseded by GLM-5.3 on 18 Aug 2026.
Benchmarks · 6 reported
| Benchmark | Score | Bar | Rank | Source |
|---|---|---|---|---|
| SWE-bench ProOpenHands, 400K context | 62.1% | #6 of 16 | GLM-5.2 model cardSelf-reported | |
| Terminal-Bench 2.1Terminus-2 harness | 81% | #11 of 17 | GLM-5.2 model cardSelf-reported | |
| GPQA Diamond | 91.2% | #13 of 27 | GLM-5.2 model cardSelf-reported | |
| Humanity’s Last Exam (no tools)no tools, text-only subset | 40.5% | #10 of 23 | GLM-5.2 model cardSelf-reported | |
| Humanity’s Last Exam (with tools)with tools, text-only subset | 54.7% | #13 of 16 | GLM-5.2 model cardSelf-reported | |
| AIME 2026 | 99.2% | – | GLM-5.2 model cardSelf-reported |
Self-reported means the lab ran the test itself. A rank appears only where several labs report the same version of a benchmark.
Similar models
Models with the most benchmarks in common with GLM-5.2, and the closest scores.
SWE-bench Pro
#6 of 16Harder multi-file engineering tasks
- GLM-5.262.1
- Qwen3.8-Max67.7
- GPT-5.558.6
- Qwen3.7-Max60.6
Terminal-Bench 2.1
#11 of 17Terminal tasks, earlier set
- GLM-5.281
- Qwen3.8-Max86.6
- Qwen3.7-Max74.5
- Kimi K388.3
- DeepSeek V4 Pro87.9
GPQA Diamond
#13 of 27Graduate-level science questions
- GLM-5.291.2
- Qwen3.8-Max92.6
- GPT-5.593.6
- Qwen3.7-Max92.4
- Kimi K393.5
- DeepSeek V4 Pro92.4
Humanity’s Last Exam (no tools)
#10 of 23Expert questions, no search or code
- GLM-5.240.5
- Qwen3.8-Max43.6
- GPT-5.541.4
- Qwen3.7-Max41.4
- Kimi K343.5
- DeepSeek V4 Pro42.7
Compare GLM-5.2 side by side
Pre-filled with the two closest models. Swap any of them.
Sources
- price: docs.z.ai/guides/overview/pricing, checked 2 Oct 2026
- context 1M, max output 128K, modalities: docs.z.ai/guides/llm/glm-5.2, checked 2 Oct 2026
- release date: docs.z.ai/release-notes/new-released, checked 2 Oct 2026
- license, benchmarks: huggingface.co/zai-org/GLM-5.2, checked 2 Oct 2026
- parameters 744B-A40B: github.com/zai-org/GLM-5, checked 2 Oct 2026
- output speed: artificialanalysis.ai/models/glm-5-2, checked 2 Oct 2026
Building on GLM-5.2?
Get the architecture right before the bill arrives.
We size caching, routing and fallbacks for your workload on a free call, and tell you where a cheaper model is good enough.