DeepSeek-V3 vs DeepSeek V3.2 Speciale
DeepSeek vs DeepSeek — benchmarks, pricing, and capabilities side by side.
- •DeepSeek V3.2 Speciale has the higher intelligence index (89.9 vs 58.1)
- •DeepSeek-V3 is cheaper ($0.23 vs $0.29 per 1M input)
- •DeepSeek V3.2 Speciale has a larger context window (164K)
| DeepSeek-V3 | DeepSeek V3.2 Speciale | |
|---|---|---|
| Intelligence index | 58.1 | 89.9 |
| Developer | DeepSeek | DeepSeek |
| Type | LLM | LLM |
| Access | Open weights | Open weights |
| Context window | 131,072 tokens | 163,840 tokens |
| Input price | $0.23 / 1M | $0.29 / 1M |
| Output price | $0.91 / 1M | $0.43 / 1M |
| Speed | 100 tok/s | — |
| Released | December 26, 2024 | December 1, 2025 |
| Parameters | 671B total / 37B active (MoE) | — |
| Input modalities | Text | Text |
| Output modalities | Text | Text |
Shared benchmarks
DeepSeek-V3
DeepSeek V3.2 Speciale
AIME 2025
26
96.7
GPQA Diamond
59.1
87.1
Humanity’s Last Exam
3.6
26.1
LiveCodeBench
37.6
89.6
MMLU-Pro
75.9
86.3