- Chat
- Reasoning
DeepSeek V4 Pro 0813
DeepSeek V4 Pro mixture-of-experts model with DSpark speculative decoding and a 1,048,576-token context.
DeepSeek model family
About this model
DeepSeek V4 Pro 0813 is a DeepSeek V4 mixture-of-experts model whose published configuration uses DSpark speculative decoding and a 1,048,576-token context. The model card recommends up to 384,000 output tokens for high and max reasoning and does not state total or activated parameter counts. The lineup is subject to change; a successor may ship in its place.
- Parameters
- —Total and activated parameter counts are not stated in the model card.
- Native context
- 1,048,5761,048,576 tokens served by Arnict.
- Output guidance
- 384,000Published guidance
Capabilities
- Chat completionsUse this model with the chat-completions API when it is marked Available.
- Native input modalitiesText input is listed in the published catalog metadata.
- Long-context input1,048,576 tokens · 1,048,576 tokens served by Arnict.
- Published output guidance384,000 tokens · Recommended maximum for high/max reasoning.
Published limits
Published model-card facts, kept separate from live availability.
| Metric | Published value | Note |
|---|---|---|
| Arnict API output limit | 384,000 tokens | Maximum output accepted by the Arnict API for this model |
| Parameters | Total and activated parameter counts are not stated in the model card. | Upstream model card |
| Context window | 1,048,576 tokens | 1,048,576 tokens served by Arnict. |
| Output guidance | 384,000 tokens recommended maximum | Recommended maximum for high/max reasoning. |
| Weights / activations | FP4 expert weights / FP8 activations | Published format |
- DSpark speculative decoding
Use deepseek/deepseek-v4-pro-0813 with an OpenAI-compatible client after this model is enabled.
Related models
DeepSeek
DeepSeek V4.1 Flash
Multimodal mixture-of-experts model: 552B backbone parameters, with 8B active during prefill and 16B during decoding.
GLM
GLM 5.3 Flash Abliterated
A derivative of GLM 5.3 Flash with modified refusal behavior for general-purpose inference.
GLM
GLM 5.3 Abliterated
A derivative of GLM 5.3 with modified refusal behavior for general-purpose inference.