OpenAI GPT pricing for enrichment

All 9 OpenAI models OpenClay supports, listed most capable first, with what 1,000 enriched rows cost on each once search fees are included.

Updated August 2026

OpenAI · 9 models · keys start with sk-

Most capable
GPT-5.6 Sol
$55.45 per 1,000 rows
Lowest cost per row
GPT-5 Nano
$10.47 per 1,000 rows

How OpenAI bills web search

OpenAI bills web search at $10 per 1,000 calls plus the search content itself as input tokens, roughly 8k per search. On a short enrichment prompt that overhead is the dominant cost, not the prompt.

Worth knowing

The GPT-5 family reasons by default. OpenClay sends an explicit low reasoning effort so a one-line lookup isn't billed for deliberation it doesn't need. Minimal effort is deliberately not used, because web search rejects it.

Every GPT model

ModelInputOutputContextPer 1k rows
GPT-5.6 Sol
gpt-5.6-sol
Hardest research questions; expensive per row, so test first
$5$30400k$55.45
GPT-5.6 Terra
gpt-5.6-terra
Strong reasoning for research that needs judgement, not just lookup
$2$12400k$28.18
GPT-5.6 Luna
gpt-5.6-luna
Newest small model, the cheapest sensible default for OpenAI
$0.2$1.2400k$11.82
GPT-5.4
gpt-5.4
Pin this to reproduce a run already validated on GPT-5.4
$2.5$15400k$32.73
GPT-5.4 Mini
gpt-5.4-mini
Step up from Mini when answers need a little more care
$0.75$4.5400k$16.82
GPT-5.1
gpt-5.1
Mature flagship if you'd rather not run a newer release
$1.25$10400k$21.71
GPT-5
gpt-5
Original GPT-5, kept for reproducibility
$1.25$10400k$21.71
GPT-5 Mini
gpt-5-mini
Well-understood workhorse for high-volume simple lookups
$0.25$2400k$12.34
GPT-5 Nano
gpt-5-nano
Cheapest tokens anywhere here, but the weakest at open-web research
$0.05$0.4400k$10.47

Using your OpenAI key

  • Get a key from OpenAI Platform. Keys start with sk-.
  • Paste it into OpenClay. It stays in browser memory for the tab, never written to disk, a database, a log or a cookie.
  • Test five rows for an exact per-row cost before running the batch.
  • If you hit rate limits, lower the parallel-requests slider. Failed rows retry automatically with exponential backoff.

Compare against other providers

No account, no card, no platform fee.

Bring your own API key and pay the model provider directly.

Try GPT on your data