What's this about?
OrangeAI detailed how DeepSeek's price increase affects his Cola Token Plan. His summary: DeepSeek used low prices to take the market and then raised them substantially — cache pricing up more than 12x, blended increases of 3–6x — forcing him to roughly triple his own plan pricing. He was candid about the consequence: Lite is now priced close to Basic, the two are similar in capability (Lite is better at Chinese writing), so users should choose deliberately rather than assume the higher tier is better value.
What exactly is a token reseller?
A token plan reseller buys inference capacity in bulk and sells subscription access, often bundling several providers. Their margin lives in the gap between wholesale and retail, which makes them structurally exposed: when the upstream provider raises prices, they can absorb it, pass it through, or exit. This is different from the grey-market token brokers in yesterday's digest — this is a legitimate business — but it sits in the same fragile position between a provider that sets prices unilaterally and customers who expect stability.
Why is this worth noticing?
- It's the concrete downstream consequence of the penetration-pricing pattern several people flagged when DeepSeek was undercutting everyone — a documented case rather than a prediction.
- The 12x cache increase is the detail that matters, because cache reads dominate coding workloads, so the blended figure understates the impact on the heaviest users.
- The advice against automatically buying the higher tier is unusually honest for a vendor, and reflects a market where tier names no longer track capability differences reliably.