DeepSeek-V4-Pro
DeepSeek-V4-Pro is an open-weights language model from DeepSeek, released April 24, 2026. It has 1.6T parameters (mixture-of-experts, 76B active). The context window is 1M tokens. On the Epoch Capabilities Index it scores 150.2, rank 69 of 256 rated models. API price: $0.66 input / $1.98 output per million tokens.
We present a preview version of DeepSeek-V4 series, including two strong Mixture-ofExperts (MoE) language models — DeepSeek-V4-Pro with 1.6T parameters (49B activated) and DeepSeek-V4-Flash with 284B parameters (13B activated) — both supporting a context length of one million tokens.
Check it against your own PC →
Can you run DeepSeek-V4-Pro locally?
Even with 32 GB of VRAM and 32 GB of RAM it is too large to run at a usable speed on a single consumer GPU. Estimates assume 32 GB of DDR5 RAM, a 16K context and llama.cpp-style GGUF quantization. Enter your exact hardware in the live app.
| GPU memory | Verdict | Best quant | VRAM | RAM | tok/s |
|---|---|---|---|---|---|
| 6 GB VRAM | Too large | – | – | – | – |
| 8 GB VRAM | Too large | – | – | – | – |
| 12 GB VRAM | Too large | – | – | – | – |
| 16 GB VRAM | Too large | – | – | – | – |
| 24 GB VRAM | Too large | – | – | – | – |
| 32 GB VRAM | Too large | – | – | – | – |
Ready-made quantizations on Hugging Face: teamblobfish/DeepSeek-V4-Pro-GGUF · nvidia/DeepSeek-V4-Pro-NVFP4 · nvidia/DeepSeek-V4-Pro-nvfp4-DSpark · RedHatAI/DeepSeek-V4-Pro-NVFP4-FP8 · amd/DeepSeek-V4-Pro-NVFP4 · mlx-community/DeepSeek-V4-Pro
Strengths by category
ModelTape's 0–100 fit on Epoch AI benchmark results.
| Category | Score | Based on |
|---|---|---|
| Reasoning | 12.6 | Chess Puzzles, Mystery Game Puzzles |
| Math | 36.9 | ProofBench, OTIS Mock AIME 2024-2025, FrontierMath-Tiers-1-3-v2-Private, FrontierMath-Tier-4-v2-Private |
| Knowledge | 75.2 | GPQA diamond, SimpleQA Verified |
Benchmark results
| Benchmark (Epoch AI) | Score |
|---|---|
| OTIS Mock AIME | 96.7% |
| GPQA Diamond | 91.5% |
| SWE-bench Verified | 77.6% |
| FrontierMath (Tiers 1–3) | 47.6% |
| SimpleQA Verified | 47.0% |
| ProofBench | 16.0% |
| Chess Puzzles | 14.8% |
| Mystery Game Puzzles | 8.6% |
| FrontierMath-Tier-4-v2-Private | 2.4% |
Links
Announcement · Hugging Face: deepseek-ai/DeepSeek-V4-Pro · Epoch AI benchmarks
Models with a similar score
GLM-5.3-Flash · Muse Spark · GLM-5.2 · Gemini 3 Pro · Claude Sonnet 4.6 · Mistral Large 4 Preview · Qwen 3.8 27B · Gemini 3 Flash
More from DeepSeek
DeepSeek V4.1 Flash 1017 · DeepSeek V4 Flash Vision Exp · DeepSeek V4.1 Flash · DeepSeek V4 Pro 0813 · DeepSeek V4 Flash 0731 · DeepSeek V4 Pro DSpark · DeepSeek V4 Flash DSpark · DeepSeek-V4-Flash