DeepSeek V3.2
by Phala · family deepseek · listed Dec 2025RTW
Hybrid-reasoning DeepSeek model with thinking and non-thinking modes, sparse attention, and tool-use
Current rates USD per 1M tokens
Input$1.00
Output$1.00
Cache read$0.50
Cache write—
Against the market
On the record
- Model id
deepseek/deepseek-v3.2- Context window
- 164K tokens
- Max output
- 64K tokens
- Input modalities
- text
- Output modalities
- text
- Reasoning
- yes
- Tool calls
- yes
- Open weights
- yes
- Knowledge cutoff
- Jul 2024
- Released
- 2025-12-01
- Last updated
- 2025-12-01
Back of the envelope
≈
Also listed at
- Alibaba Token Planfree in · free out
- Alibaba Token Plan (China)free in · free out
- TokenGo$0.217 in · $0.326 out
- DevPass (LLM Gateway)$0.26 in · $0.38 out
- LLM Gateway$0.26 in · $0.38 out
- OpenRouter$0.26 in · $0.38 out
- Kilo Gateway$0.26 in · $0.38 out
- NanoGPT$0.26 in · $0.38 out
- Meganova$0.26 in · $0.38 out
- Deep Infra$0.26 in · $0.38 out
- Poe$0.263 in · $0.384 out
- Helicone$0.27 in · $0.41 out
- Abacus$0.27 in · $0.40 out
- Alibaba (China)$0.27 in · $0.42 out
- SiliconFlow$0.27 in · $0.42 out