10,000 GPT-4 Turbo tokens cost $0.10 as input or $0.30 as output. Most workloads land in between — a 70/30 input/output split costs about $0.16. Verified Aug 28, 2026
Quick answer
For GPT-4 Turbo, 10,000 tokens cost $0.10 as input or $0.30 as output. A representative 70/30 workload is about $0.16.
Method & trust
The page prices the selected token volume using the model's listed input and output rates, then applies a 70/30 input/output split for the representative workload. Cached input, batch discounts, retries, and provider fees can change the final invoice.
Interactive calculator
10,000 in · 0 out
Input price
$10.00 / 1M tokens
Output price
$30.00 / 1M tokens
Uncached pricing. Add caching and monthly projections in the full calculator.
What 10,000 tokens buys in practice
$0.14
80% input / 20% output — short replies over conversation history.
$0.12
90% input / 10% output — long retrieved context, grounded answers.
$0.11
95% input / 5% output — full documents in, structured extraction out.
Same 10,000 tokens on other models
70/30 input/output split, uncached.
FAQ
10,000 input tokens cost $0.10; 10,000 output tokens cost $0.30. A typical 70/30 input/output mix costs about $0.16. Rates: $10.00/M input, $30.00/M output.
A chat-style request of ~4,800 tokens (4,000 in + 800 out) means 10,000 tokens is roughly 2.1 requests. For a RAG workload with ~8,500 tokens per request, about 1.2 requests.
Yes — cached input costs $5.00/M instead of $10.00/M, a 50% discount on repeated context.