DeepSeek V4 official release coming in mid-July with 2x peak-hour API pricing

- DeepSeek is a Chinese AI startup (based in Hangzhou) that has become a major rival to OpenAI and Google. It shocked the industry in early 2025 by releasing models that matched or beat GPT-4 at a fraction of the cost. - DeepSeek's models are known for being "open-weight" (the model parameters are published) and extremely efficient, using less computing power than Western competitors. - The company has faced escalating export controls from the US government, which restricts the sale of advanced chips (like Nvidia H100/B200) to China. DeepSeek's edge has been achieving high performance despite being limited to older, exportable chips. - V4 is the successor to V3 (released late 2024 / early 2025) and the R1 reasoning model. A "2x peak-hour API pricing" is a notable shift — DeepSeek has historically undercut Western AI APIs by massive margins (often 10–50x cheaper). Raising prices signals growing demand and possibly a strategy to monetize before export controls tighten further. - The mid-July 2026 target suggests this is a forward-looking report; if you're reading this after that date, the model has likely already been released.