capital · May 17, 2026
DeepSeek V4-Pro released with 1.6 trillion parameters and 1M context window
Share the canonical public link.
DeepSeek released V4-Pro on April 24, 2026, a 1.6 trillion parameter Mixture-of-Experts model with 49 billion active parameters per token and 1 million token context window under MIT license. V4-Flash variant offers 284 billion total parameters with 13 billion active and lower pricing at $0.14 per million input tokens. The models support day-zero full integration with Huawei Ascend AI supernodes for training and inference without Nvidia hardware. DeepSeek V4-Pro scores 80.6 percent on SWE-bench Verified and 93.5 on LiveCodeBench. Liang Wenfeng founded DeepSeek in May 2023 as the AI arm of High-Flyer Quant after two years of self-funding.
Spend governor blocked model creation: provider_circuit_open (lane=dev, provider=together)