DeepSeek-V4
DeepSeek-V4-Pro
DeepSeek · open-permissive mit
| Parameters (B) | 1598.84 |
|---|---|
| Architecture | moe |
| Layers / KV heads / head dim | 61 / 1 / 512 |
| Context window | 1048576 |
| Modalities | in: text; out: text |
| GGUF | available |
| Commercial use | yes |
Minimum hardware (estimate)
At Q4_K_M and 8k context, total working set ≈ 972.5 GiB. On a machine with 64 GiB system RAM and no discrete GPU: Won't run. Limiting factor: No GPU and system RAM insufficient for weights + KV.
Scores
| Benchmark | Value | Uncertainty | Run by | Source |
|---|---|---|---|---|
| GPQA Diamond | 90.9% | ±0.0205 | third-party | epoch-ai |
| SWE-bench Verified | 77.6% | ±0.0190 | third-party | epoch-ai |
| OTIS Mock AIME 2024-2025 | 96.7% | ±0.0200 | third-party | epoch-ai |
| SimpleQA Verified | 57.0% | ±0.0157 | third-party | epoch-ai |
Provenance
- identity, license, architecture, parameter count: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro (retrieved 2026-08-10)
HF: deepseek-ai/DeepSeek-V4-Pro · snapshot 2026-08-10