‹ PublicAI Index
The LLM benchmark aggregator.
GLM-4.7
Z.ai · 358B · open weights
Strongest in Mathematics (#29 of 152, on 1 of its 2 boards), weakest in Safety (#218 of 329). Above par in 9 of 18 scopes. Among the models it meets almost everywhere, it finishes behind GPT-5.5 and GPT-5.4 and ahead of Gemini 3.1 Flash-Lite and DeepSeek R1.
§ 1 · Profile
What it is good at
Bars run from 50 — the average of the models each source lists — so right of the line is above par. The middle column is the gap to whoever leads that scope.
Reasoning52.7−12.7#71/1952/5
Mathematics57.2−6.6#29/1521/2
Science52.8−10.7#69/1221/1
Reasoning48−21#75/1471/3
Human preference60.6−8.2#82/3541/1
Human preference60.6−8.2#82/3541/1
Coding48.5−17.7#96/1791/6
Agentic coding48.3−18.6#93/1731/5
Knowledge50.8−11.7#103/1971/3
Academic knowledge51.9−12#72/1231/1
Agents51.6−14.8#108/2981/6
Knowledge work53.1−17.6#74/2081/2
Professional46.2−16.2#144/1941/2
Legal49.7−16.2#91/1731/1
Medical44.9−21.6#120/1491/1
Finance43.1−19.3#131/1621/1
Safety48.6−12.8#218/3291/3
Factual grounding45.5−25.2#74/1041/1
§ 2 · Head to head
What it beats, and what beats it
The same models turn up scope after scope. Counted once: where both were placed, who finished higher. Bars run right for GLM-4.7, left for the other.
§ 3 · Sources
Where the numbers come from
5 publications, 15 figures. Every one links to the page it was read from.
LMArena Text 1441
GDPval-AA 1021
SimpleBench 47.7%
Vals · CaseLaw 54.88%Vals · LegalBench 83.36%Vals · CorpFin 46.39%Vals · TaxEval 68.77%Vals · MedQA 93.74%Vals · MedCode 32.77%Vals · MedScribe 68.63%Vals · SWE-bench Verified 69.4% (Mini-SWE-agent)Vals · GPQA Diamond 80.05%Vals · MMLU Pro 82.74%Vals · AIME 93.33%
Vectara · Factual consistency 88.3%
Badge
[](https://publicai.io/model-index/m/glm-4-7)