“DeepSeek's V4 model has a 1 million token context window, which is comparable to models from Google's Gemini, Anthropic, and OpenAI.”