Gemini 2.5 Pro Context Price Cliff
Gemini 2.5 Pro bills $1.25/M input under 200K tokens and $2.5/M over — applied to the whole prompt. This page prices the cliff for your request and the saving from staying under it.
For teams sending long prompts to Gemini Pro — computes the 200K-token price cliff, where crossing the line re-prices the whole request at a higher rate, not a smooth per-token curve.
Model prices from OpenRouter · updated 2026-07-13
01 Your prompt
≤200K: $1.25/M in · $10/M out — >200K: $2.5/M · $15/M (whole request)
02 Which side of the cliff?
Over the 200K line — the whole request bills at the high tier, +97% vs the standard rate.
The jump: $0.27 at 200K → $0.53 one token over = 2.0×
Trim 50,000 tokens (20% of the prompt) to get under 200K → save $38,500/month.
Related cost calculators
Two-tier prices hand-verified against Google's Gemini pricing page (checked 2026-07-13), re-audited quarterly: Google Gemini API pricing. Cost = prompt × (over-threshold ? high : standard) input rate + output × the matching output rate, per 1M tokens. Threshold pricing is currently Gemini-only. All math runs in your browser.
How the math works
Gemini 2.5 Pro isn't priced flat per token. Up to 200,000 prompt tokens it charges $1.25/M input and $10/M output; cross 200K and the WHOLE request re-prices at $2.5/M and $15/M — not just the tokens above the line. That's a cliff, not a ramp.
Your 250,000-token prompt is over the line, so every token bills at the high tier: $0.66/request. The same request priced entirely at the under-200K rate would be $0.33 — the cliff is adding $0.32 (97%).
Same baseline, the discontinuity: a prompt of exactly 200,000 tokens costs $0.27, but one token more (200,001) costs $0.53 — 2.0× more for a single extra token, because the higher rate applies to everything. Nowhere else on the pricing curve does one token matter that much.
The lever is obvious once you see the cliff: trim 50,000 tokens (20% of your prompt) to get back under 200K and the request drops to $0.27 — a $0.39/request saving, $38,500/month. Retrieval that keeps prompts just under the threshold pays for itself here.
This tiered mechanism is currently documented on Gemini Pro only — OpenAI prices flat per token, and Anthropic retired its old >200K premium. Prices hand-verified against Google's pricing page (checked 2026-07-13), re-audited quarterly. All math runs client-side with tested code.
Frequently asked questions
Does a bigger prompt on Gemini 2.5 Pro just cost proportionally more?
No — up to 200,000 tokens it's flat, but crossing 200K re-prices the ENTIRE request at a higher rate ($2.5/M vs $1.25/M input). So cost jumps at the threshold instead of rising smoothly. A 200,001-token prompt costs 2.0× a 200,000-token one.
What's the 200K price cliff on Gemini 2.5 Pro?
Below 200,000 prompt tokens: $1.25/M input, $10/M output. Above it: $2.5/M and $15/M — applied to the whole request. It's roughly a 2× step, so keeping prompts under the line matters more than shaving tokens anywhere else.
How do I avoid the cliff?
Keep your prompt under 200,000 tokens. Tighter retrieval (fewer/shorter chunks), summarizing old context, or dropping low-value material keeps you on the cheap tier. At this page's defaults, trimming 50,000 tokens saves $0.39/request.
Do other providers have this cliff?
Not currently. As of 2026-07-13, threshold-based context pricing is documented on Gemini Pro only. OpenAI charges flat per token regardless of prompt length, and Anthropic removed its old >200K premium (now flat across the 1M window). This calculator is scoped to Gemini for that reason.
Are these prices current?
The two-tier rates are hand-verified against Google's Gemini pricing page (checked 2026-07-13) and re-audited quarterly — they're not in the daily flat-price sync because the tiering isn't a single number. All math runs client-side with tested code.
Should Gemini 2.5 Pro worry about the cliff?
Yes — you're over the line, paying a 97% premium on the whole request ($32,250/month here). Getting back under 200K is the single biggest lever.