SIDE BY SIDE
DeepSeek V3.2 vs Hy3 preview
Each is the newest model of its family in Blossom Valley, and the two are neighbours there by blended price.
| Field | DeepSeek V3.2 | Hy3 preview |
|---|---|---|
| Provider | DeepSeek | Tencent |
| Family | DeepSeek | Tencent |
| Province | Blossom Valley | Blossom Valley |
| Context window | 164K tokens | 262K tokens |
| Maximum output | 66K tokens | 236K tokens |
| Input / 1M tokens | $0.280 | $0.180 |
| Output / 1M tokens | $0.420 | $0.600 |
| Accepts | text | text |
| Produces | text | text |
| Reasoning | Yes | Yes |
| Tool use | Yes | Yes |
| Added to OpenRouter | 2025-12-01 | 2026-04-22 |
| Model ID | deepseek/deepseek-v3.2 | tencent/hy3-preview |
Questions and answers
Which is cheaper, DeepSeek V3.2 or Hy3 preview?
It depends on the mix: Hy3 preview is cheaper for input ($0.180 vs $0.280), DeepSeek V3.2 for output ($0.420 vs $0.600) per million tokens. At three input tokens per output token, Hy3 preview costs less. On input, Hy3 preview is 36% cheaper.
Which has the larger context window?
Hy3 preview: 262K tokens against 164K for DeepSeek V3.2 (1.6× as long).
What can one do that the other cannot?
Nothing among reasoning, tool use and image input: both support reasoning and tool use.
Answers are computed from the 2026-10-24 catalog snapshot.