Model reference · open weights
DeepSeek-Pro-DSpark is an open-weight language model from nvidia, listed in the AxForge catalogue. AxForge can bring it up on EU-owned hardware for you on request — with the licence handled where one is required.
About
Model Overview Description: The NVIDIA DeepSeek-V4-Pro-nvfp4-DSpark model is the quantized version of the DeepSeek-V4-Pro model, which is a Mixture-of-Experts (MoE) language model with 1.6 trillion total parameters and 49 billion activated parameters, packaged in a single checkpoint together with DeepSeek's official DSpark speculative decoding module. For more information, please check here and here. The NVIDIA DeepSeek V4 Pro NVFP4 DSpark model is quantized with Model Optimizer. Note: DeepSeek-V4-Pro-nvfp4-DSpark is not a new model. It is the NVFP4 backbone of nvidia/DeepSeek-V4-Pro-NVFP4 with DeepSeek's official DSpark speculative decoding module attached, so a single checkpoint serves as both target and draft model. For more details on DSpark, refer to: https://github.com/deepseek-ai/DeepSpec This model is ready for commercial/non-commercial use. <br Third-Party Community Consideration This model is not owned or developed by NVIDIA. This model has been developed and built to a third-party’s requirements for this application and use case; see link to Non-NVIDIA (DeepSeek-V4-Pro-DSpark) Model Card. References - Nvidia Model Optimizer: https://github.com/NVIDIA/Model-Optimizer - DeepSeek-V4-Pro base model card - DeepSeek-V4-Pro-DSpark model card - nvidia/DeepSeek-V4-Pro-NVFP4 model card - DeepSeek-V4 technical report - DeepSeek DSpec reference implementation: https://github.com/deepseek-ai/DeepSpec License/Terms of Use: MIT Deployment Geography: Global <br Use Case: DeepSeek V4 is well-suited for advanced reasoning, agentic AI applications, tool use scenarios, and complex problem-solving in domains such as mathematics, software engineering, and enterprise AI assistants. <br Release Date: Huggingface 07/22/2026 via https://huggingface.co/nvidia/DeepSeek-V4-Pro-nvfp4-DSpark <br Model Architecture: Architecture Type: Transformers <br Network Architecture: Mixture-of-Experts (MoE) with Hybrid Attention (Compressed Sparse Attention + Heavily Compressed Attention) <br Total Parameters: 1.6 Trillion (49 Billion activated) <br Number of Draft Model Parameters: 82B in total and approximately 2.4B activated <br Draft Model Architecture: DSpark — a 3-layer semi-autoregress
Summarised from the published model card. Read the full card on the HuggingFace links below.
Specifications
| Maker | nvidia |
|---|---|
| Type | Language models |
| Parameters (lead) | 1650.5B |
| Variants | 1 |
| Runs with | Model Optimizer |
| Based on | deepseek-ai/DeepSeek-V4-Pro, deepseek-ai/DeepSeek-V4-Pro-DSpark |
| Released | 2026-07-22 |
| Popularity | 544 downloads / month |
| Licence | Open weights |
How it works
Variants
Open weights ship in several sizes and precisions. One page, all the variants — pick the one that fits your GPU. VRAM figures are estimates from model size.
| Variant | Params | Precision | VRAM | Fits 16 GB | Weights |
|---|---|---|---|---|---|
| DeepSeek-V4-Pro-nvfp4-DSpark | 1650.5B | NVFP4 | — | — | Weights ↗ |
Using it via the API
Once AxForge deploys nvidia-deepseek-pro-dspark for you, it answers on the OpenAI-compatible API — the same base URL and keys as every other model. (nvidia-deepseek-pro-dspark below is illustrative; you get the exact model name on deployment.)
$ curl -sS https://api.axforge.ai/v1/chat/completions \
-H "Authorization: Bearer $AXFORGE_API_KEY" \
-H "Content-Type: application/json" \
-d '{"model":"nvidia-deepseek-pro-dspark","messages":[{"role":"user","content":"Hello"}]}'
Licence
Open weights under mit — commercial use is permitted. Deploy it on AxForge EU hardware on request. Read the licence ↗