Gemini 2.5 Pro has a pricing trap most calculators ignore: at $1.25/$10 per million tokens under 200K input tokens, it's competitive with GPT-5.4 and Claude Sonnet. But the moment your prompt crosses 200K, the entire prompt — not just the overage — bills at the higher tier: $2.50/$15. For RAG pipelines that routinely stitch large context, this is a 73%+ surcharge nobody warned you about.
This calculator detects which tier applies based on your actual prompt length and applies the correct rate. We also model prompt caching (cached input drops to $0.10–$0.20/M) and the Batch API discount.
$1.25/M input and $10/M output for prompts under 200K tokens. Above 200K, it's $2.50/M input and $15/M output — applied to the entire prompt.
When does Gemini 2.5 Pro's pricing tier change?
At 200,000 input tokens. The new rate applies to the whole prompt, not just tokens past 200K.
What's the 2M context window?
Gemini 2.5 Pro supports up to 2 million tokens. The pricing tier still kicks in at 200K — there's no third tier.
Is Gemini 2.5 Pro cheaper than Claude Sonnet 4.6?
Under 200K input: yes, slightly. Above 200K: more expensive than Sonnet 4.6 which is flat at $3/$15 across all context lengths.
Should I switch to Gemini 2.5 Flash for long context?
Often yes. Flash is $0.30/M input flat (no tier jump) with 1M context. For RAG that exceeds 200K, Flash at flat pricing often beats Pro at tiered pricing.