No cache
$0.060000
0 writes · 0 reads · 3 ordinary inputs
Same as no cache
3 requests each bill 10,000 prefix tokens at the ordinary input rate.
OPENAI / INPUT COSTS
Prompt Caching Calculator
Same prompt. Different timing. Compare no cache, 30-minute and 12-hour caching before choosing a policy.
GPT-6.1 Sol · gpt-6.1-solResponses API · Standard · Short ContextPrices checked 2026-10-22
01 / YOUR REQUESTS
When will you reuse the prefix?
OpenAI API costs only. This tool does not estimate ChatGPT Plus / Pro subscription limits.
1–272,000 visible text tokens. At least 1,024 qualify for caching. Your entire request must stay within 272,000 input tokens for this price tier.
Use one consistent clock. Day 2 is the next day. Times are calculated chronologically, in whole minutes; daylight-saving changes and overlapping requests are not modeled.
Up to 100 sequential requests. Calculation runs in this tab; inputs are not saved or sent to an API.
02 / PREFIX INPUT COST · USD
Reusable Prefix Input Cost Comparison
Only the reusable prompt prefix input cost is compared. Dynamic input suffixes and output tokens are excluded. This is not the full API call cost or your actual bill.
Lowest in this conservative scenario: 12-hour cache.
3 requests × 10,000 prefix tokens. Each scenario starts independently with an empty cache.
30-minute cache
$0.075000
3 writes · 0 reads · 0 ordinary inputs
$0.015000 more than no cache
3 modeled writes cost $0.075000; 0 modeled reads cost $0.000000. No modeled reads offset the write premium. $0.015000 more than no cache for this prefix.
12-hour cache
$0.052000
1 writes · 2 reads · 0 ordinary inputs
$0.008000 less than no cache
1 modeled write costs $0.050000; 2 modeled reads cost $0.002000. Reads use the lower cached-input rate. $0.008000 less than no cache for this prefix.
Conservative minimum-coverage scenario, not an invoice prediction. A write modeled after the minimum TTL does not mean the real cache has expired. A simulated read does not guarantee a cache hit in a real request.
Show request billing
| Request | 30m scenario | 12h scenario |
|---|---|---|
| Day 1, 09:00 | Write $0.025000 | Write $0.050000 |
| Day 1, 13:00 | Write $0.025000 | Read $0.001000 |
| Day 1, 17:00 | Write $0.025000 | Read $0.001000 |
03 / RATES & ASSUMPTIONS
What this estimate includes
Only the same reusable, visible text prefix is priced. Assume an exact match, a single eligible explicit breakpoint, stable configuration and accounting scope, an empty starting cache, and completed sequential requests.
- Ordinary input
- $2.00 / 1M tokens
- Cached read
- $0.10 / 1M tokens
- 30m write
- $2.50 / 1M tokens
- 12h write
- $5.00 / 1M tokens
Write prices are total prices, not surcharges. Each request charges its prefix once: ordinary input, a write, or a read. No cache = tokens × requests × $2 / 1M; a cache scenario = tokens × (writes × its write rate + reads × $0.10) / 1M.
- 30m reads move the modeled coverage boundary to 30 minutes after the read.
- 12h writes establish 12 hours of minimum coverage. This conservative 12h scenario assumes no extension on reads; reads do not restart a full 12-hour period.
- At or beyond a minimum coverage boundary, the simulation assumes another write. Actual retention can be longer. These boundaries are not exact expiry times or deletion deadlines.
- The scenarios never share cache state. Switching an existing 30m entry to 12h can incur a fresh write without credit; mixed TTLs, upgrades, multiple breakpoints and prewarming are outside this MVP.
- Dynamic suffixes, hidden tokens, outputs, tools, images, regional uplifts, other service tiers, routing/concurrency effects, and account discounts are excluded. Your full request—not just this prefix—must qualify for Short Context pricing.
Prices and cache mechanisms follow the linked official documentation as checked on 2026-10-22.
This independent calculator does not connect to OpenAI or validate cache eligibility for your account. A minimum billing lifetime does not guarantee lower latency.