AiCost

OpenAI Cost Calculator — API Tokens to Dollars (2026)

Who this is for: For developers budgeting OpenAI API usage — and for anyone staring at a usage dashboard wondering what a token even costs.

GPT API pricing per token: model, input/output and cached tokens in — dollars out.

Quick answer: OpenAI API list prices (2026): GPT-5 $1.25/$10 per 1M in/out (cached $0.125); GPT-5 mini $0.25/$2; GPT-4o $2.50/$10; o3 $2/$8. Output tokens cost 8× input on GPT-5 — verbosity is the #1 bill driver. 1,000 req/day at 2k+500 tokens ≈ $7.50/day on GPT-5, $0.90 on mini.

tokens

Prompt + re-sent context

tokens

Completion length

tokens

≤ input; cached rate $0.125/M

×

Per day or month — your call

Per request
$0.0075
GPT-5 · 2000 in / 500 out
1,000 requests
$7.50
at that volume
Cached-input saving
$0.0000
set cached tokens > 0
Input (2,000 tok @ $1.25/M)$0.0025
Output (500 tok @ $10/M)$0.0050
Total × 1,000 requests$7.50
Output tokens cost 8× input on GPT-5 — trimming verbose responses is the single biggest saving. Then routing easy traffic to GPT-5 mini ($0.25/$2) cuts most stacks 60%+.
OpenAI list prices per 1M tokens, Oct 2026. Cached input bills at the cached rate (GPT-5: $0.125/M vs $1.25 fresh). Batch API is typically 50% off — not reflected here.
Core facts
GPT-5$1.25 in / $10 out /M
GPT-5 mini$0.25 / $2 /M
Cached inputGPT-5 $0.125/M (−90%)
GPT-4o$2.50 / $10 /M
8× ruleOutput ≈ 8× input price
StatusList prices collected Oct 2026 from each provider's public pricing page. Providers change tiers often — always confirm on the official page before billing decisions.

How much does the OpenAI API cost?

OpenAI bills per token at list prices per million: GPT-5 $1.25 input / $10 output; GPT-5 mini $0.25/$2; GPT-5 nano $0.05/$0.40; GPT-4o $2.50/$10; GPT-4o mini $0.15/$0.60; o3 $2/$8. Cached input runs 10–50% of fresh-input price (GPT-5: $0.125/M). Two reference points make budgets concrete: a 1,000-request day at 2k input + 500 output tokens costs ~$7.50/day on GPT-5 but ~$1.50/day on GPT-5 mini; and output tokens are 8× the price of input — trimming verbose responses is the single biggest saving. The calculator prices any token counts at exact list rates, including the cached-input discount.

Common uses

  • Pricing a feature's API cost before shipping it
  • Checking a surprise usage dashboard number against list prices
  • Comparing model tiers before committing a workload
  • Monthly budget: multiply the per-request cost by request volume

The three levers that cut an API bill

1) Cache: a stable system prompt sent as cached input runs at 10–50% of fresh price — restructure prompts so the shared prefix leads. 2) Route: send easy traffic to mini/nano tiers and reserve flagship models for hard requests — most stacks cut 60%+ with routing alone. 3) Trim output: ask for terser responses, cap max_tokens; output tokens are the expensive side of every request. Model choice is the biggest lever of the three.

Frequently Asked Questions

How much does GPT-5 cost via API?
$1.25 per 1M input tokens and $10 per 1M output — cached input $0.125/M. GPT-5 mini: $0.25/$2; nano: $0.05/$0.40. Most production chat workloads land in the mini tier's price range.
Why is my OpenAI bill so high?
Usually one of: output verbosity (output tokens cost 8× input on GPT-5), a retry loop, or context re-sent every request without caching. Pull the usage endpoint's per-model breakdown before anything else.
What are cached input tokens?
OpenAI re-uses repeated prompt prefixes at a discount — GPT-5 cached input is $0.125/M vs $1.25 fresh (90% off). Long stable system prompts + caching is the cheapest big win in API budgeting.
How do I check my actual OpenAI spend?
The usage dashboard shows per-model token usage; the costs endpoint (/v1/organization/costs) returns actual dollars if you have an admin key. Paste either endpoint's JSON into our multi-model calculator to price it.
Is the OpenAI API cheaper than ChatGPT Plus?
Depends on volume: under ~3M tokens/month of light use, API (especially mini/nano tiers) is far cheaper than $20. Heavy daily use of flagship models can exceed $20 fast — the calculator answers it for your numbers.

More API cost

AI API Cost Calculator
All providers, one calculator — plus paste any usage JSON and get the dollar figure.

Related guides