DeepSeek V4 - almost on the frontier, a fraction of the price

The 1.6T parameter claim is notable less for raw size and more for what it signals: DeepSeek is now competing on the spec sheet that enterprise buyers use to justify switching costs, not just on benchmark scores. The MIT license on a model this large is the actual headline most coverage is burying.

The cost-pressure angle here connects directly to the infrastructure spending dynamic covered in the April 17 TechCrunch tokenmaxxing cluster. If developers are already struggling with the productivity math of token-heavy workflows, a dramatically cheaper open-weights alternative at this scale changes the denominator in that calculation. Separately, Cursor's reported $50B valuation (covered here April 17) rests partly on the assumption that frontier model access remains expensive enough to justify a managed layer on top. A credible open-weights model at a fraction of frontier pricing puts quiet pressure on that assumption, even if the capability gap hasn't fully closed.

Watch whether enterprise tooling companies like Cursor publicly address DeepSeek V4-Pro compatibility within the next 60 days. If they do, it confirms the cost arbitrage is real enough to shift procurement conversations; if they stay quiet, the capability gap is probably still wide enough that price alone isn't moving buyers.