Every OpenAI API discount, side by side
| Discount | What it saves | The catch |
|---|---|---|
| Batch API | 50%, with higher rate limits | Results within 24 hours, so not for chat or agents |
| Flex processing | Batch prices on ordinary requests | Beta, some models only, slower, and busy times return 429 |
| Prompt caching | Repeated input up to 95% off, automatically | GPT-5.6 and later bill cache writes at 1.25x the input price |
| A smaller model | Luna costs a fraction of Astra | Weaker on hard tasks |
| A lower price per token | 59–75% below list on Zurelay | Chat Completions, not the Responses, Batch or Realtime APIs |
Batch API: half price within a day
The Batch API runs a file of requests at 50% off, from its own pool of higher rate limits, and returns results within 24 hours. It suits evaluations, classification, embeddings over a corpus and anything else nobody is waiting on.
Flex processing: batch prices without the batch
Set service_tier to flex and an ordinary request is billed at Batch API prices, with caching discounts on top. It's in beta for some models only, responses are slower, and when OpenAI is busy a request can fail with 429, which isn't charged, so retry with backoff and allow a long timeout. Fast mode, formerly Priority, goes the other way and costs more.
Prompt caching is automatic
OpenAI caches the repeated start of a prompt on its own, and bills cached input at a tenth of the input price on most current models and as little as 0.05x on some. GPT-5.6 and later also bill writing to the cache, at 1.25x the input price for a 30-minute lifetime. Put instructions and tool definitions first and keep them identical to hit the cache; more in prompt caching explained. Prompts over 272K tokens cost more per token on GPT models, so long context is worth trimming.
Promotional prices
OpenAI's pricing page gives GPT-5.6 Sol a promotional price until November 30, 2026. Promotions end on their date, so check the price before you plan around it.
Pick the cheapest model that does the job
Astra is the most capable GPT tier and Luna the fastest and lowest-priced. List prices per 1M tokens, with Zurelay's beside them:
| Model | Input | Output | On Zurelay (in / out) |
|---|---|---|---|
| GPT-6 Astra | $10.00 | $50.00 | $2.50 / $12.50 |
| GPT-6.1 Sol | $2.00 | $10.00 | $0.50 / $2.50 |
| GPT-6.1 Terra | $1.00 | $5.00 | $0.25 / $1.25 |
| GPT-6 Luna | $0.10 | $0.50 | $0.025 / $0.125 |
Is there an OpenAI API coupon code?
No. OpenAI's pricing and billing pages list no coupon or promo codes for the API, and ChatGPT Plus or Pro don't include API credit. Treat sites selling "OpenAI API keys" cheaply with care: shared keys get revoked, and some cheap endpoints serve a different model under GPT's name (how to check).
Pay less for every token
Zurelay sells GPT models below OpenAI's list price (59–75% below) through the Chat Completions API, so the OpenAI SDKs work after changing the base URL and the key. The lower price applies to every request, streaming included, with no batching and no waiting. On GPT-6 Astra it's below even OpenAI's Batch price:
| Model, in / out per 1M | OpenAI list | OpenAI Batch | Zurelay |
|---|---|---|---|
| GPT-6 Astra | $10.00 / $50.00 | $5.00 / $25.00 | $2.50 / $12.50 |
| GPT-6.1 Sol | $2.00 / $10.00 | $1.00 / $5.00 | $0.50 / $2.50 |
| GPT-6 Luna | $0.10 / $0.50 | $0.05 / $0.25 | $0.025 / $0.125 |
Zurelay doesn't serve the Responses, Batch, Realtime or fine-tuning APIs; for those, go direct. The OpenAI API alternative page compares the two.
Discounts read from OpenAI's Batch API, Flex processing, prompt caching and pricing docs on November 15, 2026.