DeepSeek V3.2 Speciale vs K2-V2
DeepSeek vs MBZUAI Institute of Foundation Models — benchmarks, pricing, and capabilities side by side.
- •DeepSeek V3.2 Speciale has the higher intelligence index (89.9 vs 73.6)
- •K2-V2 is cheaper ($0.00 vs $0.29 per 1M input)
| DeepSeek V3.2 Speciale | K2-V2 | |
|---|---|---|
| Intelligence index | 89.9 | 73.6 |
| Developer | DeepSeek | MBZUAI Institute of Foundation Models |
| Type | LLM | LLM |
| Access | Open weights | — |
| Context window | 163,840 tokens | — |
| Input price | $0.29 / 1M | $0.00 / 1M |
| Output price | $0.43 / 1M | $0.00 / 1M |
| Speed | — | — |
| Released | December 1, 2025 | December 5, 2025 |
| Parameters | — | — |
| Input modalities | Text | — |
| Output modalities | Text | — |
Shared benchmarks
DeepSeek V3.2 Speciale
K2-V2
AIME 2025
96.7
78.3
GPQA Diamond
87.1
68.1
Humanity’s Last Exam
26.1
9.8
LiveCodeBench
89.6
69.4
MMLU-Pro
86.3
78.6