Z.ai · Open weights
GLM 5.2 Thinking
GLM-5.2 with thinking enabled for harder long-horizon coding, autonomous agent workflows, complex engineering optimization, and real-world development tasks.
- Reasoning
- Tools
- Structured output
- Input
- Text
- Output
- Text
- Reasoning effort
- high · max
- Discovery tier
- Open weights
Pricing
USD per 1M tokens- Input
- $0.42
- Output
- $1.32
- Cache read
- $0.078
Reference prices as published by the model provider. API example
Specifications
- Context window
- 1M
- Max output
- 131.1K
- Average output (7 days)
- 1.8K tokens
- Longer than 89% of 468 models
Side by side
GLM 5.2 Thinking compared
Specifications, prices, speed and benchmark scores next to each other. Benchmarks below show every compared model too.
| GLM 5.2 ThinkingZ.aiThis model | GLM 5.2 Thinking TEEZ.aiRemove | |
|---|---|---|
| Provider | Z.ai | Z.ai |
| Discovery tier | Open weights | Open weights |
| Added | Jun 15, 2026 | Jun 16, 2026 |
| Limits | ||
| Context window | 1.05M | 1.05M |
| Max output | 131.07K | 131.07K |
| Average output | 1.8K tokens | 2K tokens |
| Pricing, USD per 1M tokens | ||
| Input | $0.42Best | $1.40 |
| Output | $1.32Best | $4.40 |
| Cache read | $0.078Best | $0.70 |
| Performance | ||
| Throughput | 108.0 tok/s | — |
| Time to first token | 2.0 s | — |
| Request success (24h) | 99.7%Best | 87.5% |
| Benchmark indexes | ||
| Intelligence Index | 33.7 | 33.7 |
| Coding Index | 68.8 | 68.8 |
| Capabilities | ||
| Reasoning | Yes | Yes |
| Tools | Yes | Yes |
| Structured output | Yes | Yes |
| Reasoning effort | high, max | low, medium, high, xhigh, max |
Public evaluations
Benchmarks
Independent evaluation results, with how each score ranks among the models in the Noumen catalog.
- GLM 5.2 Thinking
- GLM 5.2 Thinking TEE
Intelligence Index
- GLM 5.2 Thinking33.7
- GLM 5.2 Thinking TEE33.7
GLM 5.2 Thinking tested at max reasoning effort
GLM 5.2 Thinking: Better than 76% of 326 scored models
Indexes
Coding Index
Composite of the coding evaluations
- GLM 5.2 Thinking: 68.8Top
- GLM 5.2 Thinking TEE: 68.8Top
Agentic work
AutomationBench-AA
Workflow automation, penalized for guardrail violations
- GLM 5.2 Thinking: 28.4%Top
- GLM 5.2 Thinking TEE: 28.4%Top
AutomationBench-AA tasks completed
Workflows finished end to end without a guardrail violation
- GLM 5.2 Thinking: 6.2%Top
- GLM 5.2 Thinking TEE: 6.2%Top
AA-Briefcasev1.1
Agentic knowledge work, rated head to head
- GLM 5.2 Thinking: 1,227 EloTop
- GLM 5.2 Thinking TEE: 1,227 EloTop
GDPval-AAv2.1
Economically valuable occupational tasks, rated head to head
- GLM 5.2 Thinking: 1,373 EloTop
- GLM 5.2 Thinking TEE: 1,373 EloTop
Document reasoning
GDP.pdf
Professional reasoning over PDF documents, all-pass rate
- GLM 5.2 Thinking: 10.4%Top
- GLM 5.2 Thinking TEE: 10.4%Top
AA-LCRv1.1
Reasoning across long documents
- GLM 5.2 Thinking: 78.3%Top
- GLM 5.2 Thinking TEE: 78.3%Top
MLCR-AA
Long-context reasoning over medical records
- GLM 5.2 Thinking: 12.8%Top
- GLM 5.2 Thinking TEE: 12.8%Top
Reasoning and knowledge
Humanity’s Last Exam
Expert-level questions across academic fields
- GLM 5.2 Thinking: 41.1%Top
- GLM 5.2 Thinking TEE: 41.1%Top
GPQA Diamond
Graduate-level science questions
- GLM 5.2 Thinking: 89.5%Top
- GLM 5.2 Thinking TEE: 89.5%Top
IFBench
Following precise output instructions
- GLM 5.2 Thinking: 73.3%Top
- GLM 5.2 Thinking TEE: 73.3%Top
Coding
Terminal-Benchv4.0
Practical software engineering in a terminal
- GLM 5.2 Thinking: 1.0%Top
- GLM 5.2 Thinking TEE: 1.0%Top
SciCode
Scientific computing in Python
- GLM 5.2 Thinking: 51.2%Top
- GLM 5.2 Thinking TEE: 51.2%Top
Legacy evaluations
Terminal-Bench Hard
Earlier terminal task suite
- GLM 5.2 Thinking: 50.8%Top
- GLM 5.2 Thinking TEE: 50.8%Top
τ²-Bench Telecom
Tool use in a customer-support simulation
- GLM 5.2 Thinking: 99.1%Top
- GLM 5.2 Thinking TEE: 99.1%Top
AA-LCR (unversioned)
Long-context reasoning, earlier grading
- GLM 5.2 Thinking: 78.3%Top
- GLM 5.2 Thinking TEE: 78.3%Top
Scores from Artificial Analysis, updated Oct 26, 2026. GLM 5.2 Thinking was tested at max reasoning effort; compare scores measured at the same effort.Artificial Analysis
- GLM 5.2 Thinking
- GLM 5.2 Thinking TEE
| Leaderboard | GLM 5.2 Thinking | GLM 5.2 Thinking TEE |
|---|---|---|
| Industry Legal And Government | #18 · 1491.1 | Not in top 20 |
| Spanish | #16 · 1478.5 | Not in top 20 |
Ranks among Noumen catalog models on LMArena leaderboards. Only top-20 placements are listed.LMArena
- GLM 5.2 Thinking
- GLM 5.2 Thinking TEE
| Leaderboard | GLM 5.2 Thinking | GLM 5.2 Thinking TEE |
|---|---|---|
| FrontierMath v2 (Tiers 1–3) | #7 · 59.2% | Not in top 20 |
| SWE-bench Verified | #3 · 78.7% | Not in top 20 |
| GPQA Diamond | #6 · 91.9% | Not in top 20 |
Ranks among Noumen catalog models on Epoch AI leaderboards. Only top-20 placements are listed.Epoch AI
Live routing data
Performance and reliability
Recent averages across requests routed for this model. Your own speed depends on prompt length, reasoning effort and load.
- Throughput
- 108.0 tok/s
- Faster than 66% of 253 models
- Time to first token
- 2.0 s
- Quicker than 78% of 253 models
- Prompt cache hits
- 59.1%
- About $0.22 per 1M input tokens with caching
- Request success (24h)
- 99.7%
- Reliable
Cache hits are the share of input tokens recently served from a prompt cache across all requests. They are not a forecast of your own hit rate.
- Healthy (98% or more)
- Degraded (90–98%)
- Down (under 90%)
- No requests
Developers
Use it through the API
Noumen’s API is OpenAI-compatible: any Chat Completions client works with a base URL and a key.
curl https://www.noumen.xyz/api/v1/chat/completions \
-H "Authorization: Bearer $NOUMEN_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "z-ai/glm-5.2:thinking",
"reasoning_effort": "high",
"messages": [{ "role": "user", "content": "Explain this stack trace." }]
}'import os
from openai import OpenAI
client = OpenAI(base_url="https://www.noumen.xyz/api/v1", api_key=os.environ["NOUMEN_API_KEY"])
reply = client.chat.completions.create(
model="z-ai/glm-5.2:thinking",
messages=[{"role": "user", "content": "Explain this stack trace."}],
)
print(reply.choices[0].message.content)Z.ai
Related models
Other Z.ai models, starting with the same family.