AI models
GLM-5.3 vs DeepSeek V4 Pro
Same criteria, side by side. Where the gap is small, it may be measurement noise.
Also searched as
GLM-5 vs DeepSeek V4
Direct answer: GPQA Diamond: GLM-5.3 90.9%, DeepSeek V4 Pro 91.7% (Epoch AI). OTIS Mock AIME: GLM-5.3 91.1%, DeepSeek V4 Pro 98.6% (Epoch AI). FrontierMath (tiers 1 to 3): GLM-5.3 68.8%, DeepSeek V4 Pro 64.6% (Epoch AI). SimpleQA Verified: GLM-5.3 41%, DeepSeek V4 Pro 52.9% (Epoch AI). GPQA Diamond (OpenRouter): GLM-5.3 87.9%, DeepSeek V4 Pro 91.9% (OpenRouter Benchmarks). τ-bench Verified (airline): GLM-5.3 78.8%, DeepSeek V4 Pro 78.6% (OpenRouter Benchmarks).
| Metric | GLM-5.3 | DeepSeek V4 Pro | Source |
|---|---|---|---|
| Weights ?what it is | open | open | no data |
| Input US$/1M ? | US$ 0.07 | US$ 0.96 | no data |
| Output US$/1M ? | US$ 1.40 | US$ 1.91 | no data |
| Context ? | 1,048,576 | 1,048,576 | no data |
| GPQA Diamond ? | 90.9% | 91.7% | Epoch AI |
| OTIS Mock AIME ? | 91.1% | 98.6% | Epoch AI |
| FrontierMath (tiers 1 to 3) ? | 68.8% | 64.6% | Epoch AI |
| SimpleQA Verified ? | 41% | 52.9% | Epoch AI |
| SWE-bench Verified ? | no data | 77.6% | Epoch AI |
| GPQA Diamond (OpenRouter) ? | 87.9% | 91.9% | OpenRouter Benchmarks |
| τ-bench Verified (airline) ? | 78.8% | 78.6% | OpenRouter Benchmarks |
Compare with
See lab: Zhipu AI (Z.ai) · DeepSeek
Sources
- Epoch AI, CC BY 4.0, latest version we checked.
- OpenRouter Benchmarks, CC BY 4.0, latest version we checked.
- OpenRouter, dados públicos da API; exports de benchmarks e rankings em CC BY 4.0, latest version we checked.
Dates come from the sources: OpenRouter listing or Hugging Face repo creation, which may differ from the official launch. Family and version were grouped by rule from the name.