Z.ai: glm-5.3
21 benchmarks
Leaderboard position
#31
Overall rank (of 1122 models)
2729
Elo rating
21
Benchmarks
Top 2.8%
Percentile
In Z.ai (of 18 models)
Above: OpenAI: gpt-5.6-terra #30
Below: OpenAI: GPT-5.5 Pro #32
Benchmark highlights
Best benchmark
GDPval-AA v2
1769
Worst benchmark
Agent Arena - Net Improvement (25.09.2026)
2.86
Nearest competitor from another company
The model from another company with the most similar benchmark scores.
appeared as nearest in 3 of 21 benchmarks
3/21
14.3%
| Model | Votes | % of benchmarks |
|---|---|---|
| Anthropic: Claude Fable 5 | 3 | 14.3% |
| DeepSeek: deepseek-v4.1-flash | 2 | 9.5% |
| DeepSeek: deepseek-v4-flash-vision-exp | 2 | 9.5% |
| SpaceXAI: grok-4.6 | 2 | 9.5% |
| DeepSeek: deepseek-v4.1-flash-1017 | 2 | 9.5% |
| Moonshot AI: kimi-k3 | 1 | 4.8% |
| OpenAI: GPT-5.5 | 1 | 4.8% |
| inclusionAI: ling-3.1-flash | 1 | 4.8% |
| Anthropic: Claude Opus 4.8 | 1 | 4.8% |
| OpenAI: gpt-5.5-cyber | 1 | 4.8% |
Benchmarks
GDPval-AA v2
1769
StoryBench
1254
CommunicationBench
1007
Human-Likeness
996
Terminal-Bench v2.1
88.2
CyberGym
84.5
FrontierSWE
78.1
Toolathlon Verified
73
DeepSWE 1.1
66.9
Humanity's Last Exam (with tools)
62.5
NL2Repo
58
ExploitBench
54.4
Artificial Analysis Intelligence Index v4.3
45
PostTrainBench
39.8
Artificial Analysis Intelligence Index v4.3.2
35
Artifical Analysis Intelligence Index v5.0
32
Artificial Analysis Intelligence Index v5.0.1
32
Agents' Last Exam (Pass Rate) (Official Agent)
28.5
Terminal-Bench v3.0
28.3
Program Bench
19
Agent Arena - Net Improvement (25.09.2026)
2.86
Predicted scores
Estimated from 25 similar models and the global Elo rating. These are projections, not measured results.