Open MoE flagship with million-token context for coding and long agent runs
Details
| Field | Value |
|---|---|
| id | deepseek-v4-pro |
| family | deepseek-thinking |
| modality | text->text |
| context_length | 1000000 |
| max_output_tokens | 131072 |
| knowledge_cutoff | 2025-05 |
| release_date | 2026-04-24 |
| is_open_weights | yes |
| supports_tool_call | yes |
| supports_reasoning | yes |
| supports_structured_output | yes |
| supports_attachment | no |
| input_modalities | text |
| output_modalities | text |