Hybrid-reasoning DeepSeek model with thinking and non-thinking modes, sparse attention, and tool-use
Details
| Field | Value |
|---|---|
| id | deepseek-v3-2 |
| family | deepseek |
| modality | text->text |
| context_length | 128000 |
| max_output_tokens | 32768 |
| knowledge_cutoff | 2024-07 |
| release_date | 2025-12-01 |
| is_open_weights | yes |
| supports_tool_call | yes |
| supports_reasoning | yes |
| supports_structured_output | yes |
| supports_attachment | no |
| input_modalities | text |
| output_modalities | text |