DeepSeek has announced a 75% promotional discount on its newly released V4-Pro model for developers until 5 May 2026, alongside a permanent reduction of API cache-hit prices to one-tenth of previous levels across its entire suite. At full price, V4-Pro already undercuts OpenAI GPT-5.5, Google Gemini 3.1 Pro, and Anthropic Claude Opus 4.7 on per-token costs. The promotional input price drops to roughly $0.036 per million tokens. V4-Pro is a 1.6 trillion parameter mixture-of-experts open-weight model with a 1 million-token context window, trained on Huawei Ascend 950 chips rather than Nvidia GPUs. The move intensifies an ongoing AI pricing war and arrives amid US government accusations of Chinese firms conducting industrial-scale distillation of American AI models.
Sort: