In partnership with
DeepSeek is back. And the price is absurd.
DeepSeek quietly dropped two new models today with no prior announcement: DeepSeek V4-Pro and V4-Flash. Both come with a one million token context window and are fully open source under the MIT license. The Pro version is one of the largest open-weight models ever released. You can download it, run it yourself, and pay nothing.
V4-Pro scores 80.6% on SWE-bench Verified, within 0.2 points of Claude Opus 4.6, and costs $3.48 per million output tokens compared to Claude's $25. That is a 7x price gap at near-identical coding performance. The Flash model is even cheaper at $0.28 per million tokens, making it one of the most affordable models available from any lab anywhere in the world.
It is not quite state of the art. On Humanity's Last Exam, which tests expert-level reasoning across disciplines, V4-Pro scores 37.7% compared to Claude at 40% and GPT-5.4 at 39.8%. Early users describe it as a new favorite for creative writing, highly capable for coding, but not quite matching the very best American models on the hardest reasoning tasks. DeepSeek acknowledges the gap directly in their own technical writeup.
The story here is not that DeepSeek beat everyone. It is that they got this close, for this little, and gave the weights away for free. That combination is what keeps American labs up at night. A year ago the gap between Chinese and American AI felt decisive. Today it looks like a rounding error, and the whole thing is open source for anyone to download and build on.
Google just wrote Anthropic a $40 billion check.
Alphabet, the parent company of Google, announced today it is investing up to $40 billion in Anthropic, along with at least 5 gigawatts of computing power over five years. That is an almost incomprehensible amount of compute, equivalent to powering a small city, dedicated entirely to running Anthropic's models.
The context that makes this even more remarkable: Amazon committed up to $25 billion to Anthropic just four days ago. That means Anthropic has received commitments of up to $65 billion from two of the world's largest tech companies in under a week.
But here is the part worth sitting with. Both Google and Amazon build and sell their own competing AI models. Google has Gemini. Amazon has Nova. Neither of them is betting that Anthropic wins the AI race. What they are both doing is making sure Anthropic runs on their computing infrastructure, because whoever's servers power the winning AI company collects enormous fees for every query, every API call, every product built on top of it. The investment is really a way of locking in a massive future customer.
As one person put it online: Google just wrote a $40 billion check to the company that makes the AI that is supposed to replace Google Search, and somehow that sent Google past a $4 trillion market cap. The AI industry has entered a phase where the logic is genuinely hard to follow, and the numbers involved have stopped feeling real.
Arnold Schwarzenegger has a newsletter.
Yeah. That Arnold Schwarzenegger.
So do Codie Sanchez, Scott Galloway, Colin & Samir, Shaan Puri, and Jay Shetty. And none of them are doing it for fun. They're doing it because a list you own compounds in ways that social media never will.
beehiiv is where they built it. You can start yours for 30% off your first 3 months with code PLATFORM30. Start building today.
Claude Code got a refresh on web and mobile.
Anthropic updated Claude Code on the web today with a new look that matches the desktop app redesign, faster responses, and better reliability. You can now start a session on the web or mobile and continue it in your terminal right where you left off, with full history and the right code branch already ready to go.
Small update in the context of everything else today, but Claude Code keeps getting more polished with every week that passes.