DeepSeek has released its V4 model, sending shockwaves through the AI industry with dramatically reduced pricing. The company's Flash tier now costs just $0.14 per million input tokens and $0.28 per million output tokens—a fraction of the $1.74/$3.48 charged by its Pro tier. This price drop, reported to be 99% cheaper than some competitors like Opus, has sparked debate about the future of AI model pricing.
Other highlights from the video include:
- Claude now requires ID verification, raising privacy concerns.
- An "intelligence crash" is discussed, where many models appear to be getting "dumber."
- OpenAI Agents SDK 2.0 may signal the end of traditional pull requests.
- MiniMax M2.7 is noted for bugs and fine-tuning pitfalls.
The video also emphasizes that DeepSeek V4 is open source, potentially democratizing access to high-performance AI.