DeepSeek V3.2 is 2.0× cheaper per million tokens (blended). DeepSeek V3.2 has the larger context window (163,840 vs 128,000 tokens).
DeepSeek V3 vs DeepSeek V3.2
Live catalog fields. Best-in-row is highlighted.
DeepSeek V3open
DeepSeek · deepseek__deepseek-chat
DeepSeek V3.2open
DeepSeek · deepseek__deepseek-v3.2
Identity
| Field | deepseek__deepseek-chat | deepseek__deepseek-v3.2 |
|---|---|---|
| Provider | DeepSeek | DeepSeek |
| Slug | deepseek__deepseek-chat | deepseek__deepseek-v3.2 |
| Status | live | live |
| Open weights | Yes | Yes |
| License | — | — |
| HuggingFace | deepseek-ai/DeepSeek-V3 | deepseek-ai/DeepSeek-V3.2 |
Quality
| Field | deepseek__deepseek-chat | deepseek__deepseek-v3.2 |
|---|---|---|
| Quality index (0–100) | — | — |
Cost
| Field | deepseek__deepseek-chat | deepseek__deepseek-v3.2 |
|---|---|---|
| Input $ / 1M tokens | $0.26 | $0.26 |
| Cached input $ / 1M | — | $0.13 |
| Output $ / 1M tokens | $1.03 | $0.38 |
Context
| Field | deepseek__deepseek-chat | deepseek__deepseek-v3.2 |
|---|---|---|
| Max context tokens | 128,000 | 163,840 |
| Max output tokens | 16,000 | 147,456 |
Modalities
| Field | deepseek__deepseek-chat | deepseek__deepseek-v3.2 |
|---|---|---|
| Input | text | text |
| Output | text | text |