deepseek/deepseek-v4-proDeepSeek-V4-Pro is DeepSeek’s open-weights AI model, released Apr 24, 2026. It scores 150 on the Epoch Capabilities Index, costs $1/M input and $1.50/M output tokens, and handles a 1M-token context window. Its strongest use case is agentic & tools, where it ranks in the 91st percentile of the models Model Beat tracks.
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window.
Scores on standardized evaluations. Higher is better — and rank shows where DeepSeek-V4-Pro lands among all models tracked on Model Beat.
Graduate-level scientific reasoning
Frontier of human expert knowledge
Common-sense trick questions
Factual accuracy & hallucination
Novel ML problem-solving
Scientific research coding
Real GitHub issue resolution
Human-rated web development
Olympiad-qualifier math
Tool-agent-user reliability
DeepSeek V4 Pro is a large-scale Mixture-of-Experts model from DeepSeek with 1.6T total parameters and 49B activated parameters, supporting a 1M-token context window.
DeepSeek-V4-Pro is an AI model developed by DeepSeek, part of the DeepSeek family, released Apr 24, 2026.
DeepSeek-V4-Pro is served from $1 per million input tokens and $1.50 per million output tokens (cheapest credible provider on OpenRouter: SiliconFlow).
DeepSeek-V4-Pro supports a context window of up to 1M tokens.
On standardized evaluations tracked by Epoch AI, DeepSeek-V4-Pro scores 96.2% on τ²-bench, 77.6% on SWE-bench Verified, 90.9% on GPQA Diamond.
DeepSeek-V4-Pro is released with open weights (Open weights (unrestricted)), so the model can be downloaded and self-hosted.
Benchmark data from Epoch AI (CC BY) and Artificial Analysis; pricing, specs & descriptions from OpenRouter. Data updated Nov 12, 2026.