OpenAI's Pro tier hit its floor at $15 — and has doubled in the ten months since

OpenAI has sold a “think harder” tier since March 2025, and the price of that tier has not moved in one direction. Reading its model cards end to end gives a curve that falls off a cliff and then climbs steadily back. o1-pro opened the line in the API in March 2025 at $150 per million input tokens and $600 per million output. o3-pro replaced it in June 2025 at $20 and $80 — an 87 per cent cut on both sides in three months. GPT-5 Pro arrived on 6 October 2025 cheaper still on input, $15, but dearer on output at $120, so the two halves of the price bottomed out at different moments. From there the direction reverses: GPT-5.2 Pro on 11 December 2025 at $21 and $168, GPT-5.4 Pro on 5 March 2026 at $30 and $180, and GPT-5.5 Pro on 23 April 2026 holding those same figures — the first pause after three consecutive rises. Measured from the floor, input has gone from $15 to $30 and output from $80 to $180. That is a doubling and a 125 per cent increase respectively, inside ten months, on a product line whose public story is that inference keeps getting cheaper. Two honest caveats. The models are not the same product: GPT-5.5 Pro carries a 1,050,000-token context window against 400,000 for GPT-5 Pro, a knowledge cutoff eight months fresher, and reasoning effort that defaults to high. And the cheap end of the catalogue did keep falling — GPT-5.4 nano sells at $0.20 per million input tokens, which puts a single generation of one family across a 150-fold price range from nano to Pro. What the curve does show is that the top of OpenAI's range is not subject to the same gravity as the bottom of it, and that a buyer who standardised on the Pro tier at its cheapest point is now paying twice as much for the equivalent seat.