New Delhi, May 18 -- When DeepSeek released the preview of its V4 model on 24 April 2026, the contrast with the rest of the AI industry could not have been starker (tinyurl.com/4d6uauvn).

The Pro variant is priced below Claude Opus 4.7, GPT-5.5 and Gemini 3.1 Pro, and the Flash variant undercuts every smaller frontier model. Further, it is open-source and runs on domestic Chinese chips from Huawei and Cambricon, showing the world that Nvidia's chokehold on the AI hardware space is not absolute.

In the same week, Anthropic users hit Claude Code usage caps faster than expected, OpenAI scaled back Sora to redirect compute towards core services and Datadog reported nearly 60% of AI failures in production now trace back to capacity limits (t...