DeepSeek V3.2 vs GPT-5
Head to head on 19 shared benchmarks, with price, context window, and release dates.
Benchmark scores
76evals · sort by any column, ⤢ to expand fullscreen
| Model | |||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| GPT-5 OpenAI | 88.0% | 48.3% | 61.1% | 1623elo | 72.8 | 32.4% | 68.6% | 6.6% | 45.0% | 83.4% | 54.3% | 82.0% | 18.0% | 37.8% | 74.4% | 73.4% | 12.9% | 20.1% | 0.0% | 46.7% | - | 91.1% | 94.6% | 0.0% | 0.15 | 0.0% | 68.1% | 4.58 | - | - | 57.7% | 80.0% | 80.0% | 91.7% | 0.0% | 88.9% | 83.3% | 100.0% | 48.9% | - | 58.2% | - | - | - | - | - | - | - | 37.5% | 0.52 | 1.62 | 100.0% | 86.1% | - | 0.0% | 92.7% | 92.7% | 49.6% | 83.7% | 93.3% | 10.0% | 0.0% | 65.5% | 60.0% | 0.26 | 0.0% | 43.7% | 74.9% | - | 2.34 | - | 87.5% | 0.86 | 54.8% | 90.0% | 90.0% | 56.7% |
| DeepSeek V3.2 DeepSeek | 74.2% | 59.0% | 47.9% | 1511elo | 66.8 | 22.1% | 75.1% | 11.2% | 49.0% | 68.0% | 59.3% | 83.7% | 8.0% | 38.7% | 70.0% | 68.2% | 32.6% | 5.1% | 78.9% | - | 74.7% | - | - | - | - | - | - | - | 0.0% | 16.0% | - | - | - | - | - | - | - | - | - | 31.3% | - | 65.9% | 64.6% | 50.0% | 48.2% | 70.4% | 85.0% | 77.2% | - | - | - | - | - | 57.5% | - | - | - | - | - | - | - | - | - | - | - | - | - | - | 59.0% | - | 65 | - | - | - | - | - | 51.7% |
2 / 2 models
| Attribute | DeepSeek V3.2 DeepSeek | GPT-5 OpenAI |
|---|---|---|
| Description | DeepSeek V3.2 is an AI model from DeepSeek, released with open weights. | OpenAI's August 2025 unified frontier model that auto-routes between a fast model and a deeper "thinking" variant. |
| Family | Deepseek | GPT 5 |
| Params | - | Undisclosed (closed) |
| Open weights | open | closed |
| License | mit | Proprietary |
| Released | 1 Dec 2025 | 7 Aug 2025 |
| Context | 163,840 | 400,000 |
| Input $/1M | $0.28 | - |
| Output $/1M | $0.42 | - |
| Cache read $/1M | $0.130 | $0.125 |
| Cache write $/1M | - | - |
| Speed | 0 tok/s | 0 tok/s |
| Latency | 0.00 s | 0.00 s |
| Modalities | text | text, image, file |
| Providers | - | openai |
| Publisher | DeepSeek | OpenAI |
| Reported scores | Aider Polyglot Benchmark74.2% (Aider) MathArena57.47% GPQA Diamond75.1% (AA) Humanity's Last Exam (HLE)11.2% (AA) |
FAQ
- Is DeepSeek V3.2 or GPT-5 better?
- Across the 19 benchmarks both models report on Sophon, GPT-5 leads on 10 of them. Which one is "better" depends on the benchmark - the table above shows every shared score.
- How do DeepSeek V3.2 and GPT-5 compare on GPQA Diamond?
- DeepSeek V3.2 scores 75.1% and GPT-5 scores 68.6% on GPQA Diamond.
- When were DeepSeek V3.2 and GPT-5 released?
- DeepSeek V3.2 was released 1 Dec 2025 by DeepSeek. GPT-5 was released 7 Aug 2025 by OpenAI.
Comparing something else? Build your own side-by-side.