OpenAI API credits in 2026: free credits, prepaid billing and tiers
Whether new OpenAI accounts get free API credits in 2026, how prepaid billing and usage tiers work, and the Batch, cached input and Flex discounts.
OpenAI doesn't publish any free trial credit for new API accounts on its pricing or rate limits pages. New organizations in approved geographies get a Free usage tier capped at $100 per month, and buying $5 of prepaid credits moves them to the Build tier.
If you opened an OpenAI API account hoping for a free credit and found $0.00, that’s not a glitch. OpenAI API credits are prepaid dollars you buy before you call the API, and nothing official promises any of them for free. What OpenAI does publish is a Free usage tier with a $100 monthly ceiling, three paid tiers that upgrade automatically at $5, $100 and $500 of total purchases, and three discounts that routinely cut a bill in half or better. Everything here was checked on OpenAI’s own developer pages on November 2, 2026.
Whether you get anything free
Search forums and you’ll find people asking why their new account shows no credit. The answer is in the usage tiers. OpenAI lists four of them and ties the first paid one to credit purchases rather than a grant. The Free tier is for users in an approved geography and allows up to $100 of usage per month. There’s no sign-up credit on the price list, and the spend limits guide describes a credit_balance_exhausted error for organizations that have run out of prepaid credits, which only makes sense if the balance starts with a purchase.
We’re not repeating the $5 and $18 figures from older third-party posts, because nothing official confirms them. If your dashboard shows a credit grant, that’s a promotion or partner benefit tied to your account, not a standard offer. Our OpenAI credits page lists any promotion we can verify.
One related point for people who searched for a ChatGPT Plus discount: ChatGPT Plus is a consumer subscription, billed separately from the API, and it doesn’t include API credits. Nothing in this guide applies to it.
How prepaid billing works
The API runs on a credit balance. You add funds on the billing page, each request deducts its token cost, and when the balance hits zero requests fail with credit_balance_exhausted until you top up. The fix is simply to add credits, though enforcement isn’t instantaneous, so a burst of traffic can push recorded spend slightly past a limit before requests stop.
Two settings on the organization limits page deserve attention on day one. A spend alert sends a notification when monthly spend reaches an amount you choose and lets traffic continue. A hard spend limit, enabled with the “Enforce a hard limit” toggle, returns a 429 error with the code organization_spend_limit_exceeded once the monthly figure is reached, and the same control exists per project with project_spend_limit_exceeded. Traffic resumes when you raise the limit or when the monthly cycle resets.
Auto-recharge, minimum purchase amounts and credit expiry are set on your billing page and explained in OpenAI’s help center. Read those settings before you rely on them, because an auto-recharge that’s switched on will keep buying credits as a runaway script spends them. We’d leave it off until a project has a few weeks of predictable spend behind it.
Usage tiers and how they upgrade
Older guides describe numbered tiers 1 through 5 with waiting periods after your first payment. That’s gone. There are now four named tiers, and the only qualification is cumulative credit purchases.
| Tier | Qualification | Monthly usage limit |
|---|---|---|
| Free | Located in an approved geography | $100 |
| Build | $5 in total credit purchases | $500 |
| Launch | $100 in total credit purchases | $5,000 |
| Grow | $500 in total credit purchases | $200,000 |
Your organization’s tier upgrades automatically as total purchases reach each threshold, and higher tiers generally come with higher rate limits across models. There’s no waiting period in the table. The monthly usage limit is the amount OpenAI approves you to spend; it’s separate from the spend limits you set yourself, and crossing it produces a distinct organization_usage_limit_exceeded error. You raise that one by requesting a higher approved usage limit, not by buying more credits.
In practice, a $500 purchase on day one lands you in Grow with a $200,000 monthly ceiling, which is more than most startups will ever need before they talk to sales.
What the tokens cost
OpenAI lists standard, Batch and Flex rates per 1 million tokens. The GPT-5 family is the one to know.
| Model | Input | Cached input | Output | Batch input | Batch output |
|---|---|---|---|---|---|
| gpt-5 | $1.25 | $0.125 | $10.00 | $0.625 | $5.00 |
| gpt-5-mini | $0.25 | $0.025 | $2.00 | $0.125 | $1.00 |
| gpt-5-nano | $0.05 | $0.005 | $0.40 | $0.025 | $0.20 |
| gpt-4.1 | $2.00 | $0.50 | $8.00 | $1.00 | $4.00 |
| gpt-4o-mini | $0.15 | $0.075 | $0.60 | $0.075 | $0.30 |
There’s also a 10% uplift on regional data-residency endpoints for models released on or after March 5, 2026, which doesn’t touch the GPT-5 family above. For a side-by-side with Claude pricing, see our OpenAI vs Anthropic comparison.
Batch API, 50% off for a 24-hour window
The Batch API takes 50% off compared to synchronous requests. You upload a JSONL file of requests, OpenAI processes them within 24 hours, often faster, and you download the results. A single batch can hold up to 50,000 requests or 200 MB, you can create up to 2,000 batches per hour, and there’s currently no output-token limit.
The math is simple: gpt-5 drops from $1.25 to $0.625 per 1M input tokens and from $10.00 to $5.00 output. Any job that can wait, such as nightly classification, embeddings, evaluation runs or document summarization, should run through Batch by default. Each model also has a cap on queued prompt tokens, shown on your platform settings page.
Cached input, up to 95% off repeated prompts
Prompt caching is on by default for supported models. On the GPT-5 family cached input costs 10% of the standard input rate, so a cached gpt-5 token costs $0.125 instead of $1.25 per million. The discount reaches up to 95% on newer models, with cache reads at 0.1 times the input rate on GPT-5.6 and later and 0.05 times on the GPT-6.1 models.
The rules that decide whether you get the discount are specific. A prefix must be at least 1,024 visible input tokens before it can be cached on GPT-5.6 and later. The default cache lifetime is 30 minutes, and a 12-hour option costs 2.5 times the uncached input rate to write instead of 1.25 times. Put your static content first, your system prompt, tool definitions and reference documents, and keep the variable user message at the end so the prefix stays identical across calls.
Flex processing, Batch prices on live requests
Flex processing is a beta service tier that bills tokens at Batch API rates while still returning a response in the same call. Flex is listed for gpt-5, gpt-5-mini and gpt-5-nano at exactly the Batch figures, so gpt-5 costs $0.625 input and $5.00 output. You enable it by setting service_tier to flex on a Responses or Chat Completions request.
The trade is reliability. Responses are slower, OpenAI recommends raising your client timeout to around 15 minutes, and requests can fail with a 429 Resource Unavailable error during busy periods. You aren’t charged for those failures. If your workload tolerates retries with backoff, Flex is the cheapest way to get a response today rather than tomorrow.
Startup credits and partner programs
We don’t quote a credit amount for OpenAI’s own startup program, because there’s no official figure we can stand behind. Third-party listings range from $2,500 to six figures and disagree with each other, which is a good sign that the amounts depend on the partner or investor that refers you. Ask your accelerator or VC whether they hold an OpenAI credit allocation before you apply anywhere. Our startup credits hub lists the programs we can verify, and the OpenAI review covers the platform beyond pricing.
Spending less, in order of impact
- Route anything that can wait through the Batch API for a flat 50% off.
- Structure prompts so the first 1,024 tokens never change, and let caching take 90% off that prefix.
- Use gpt-5-nano or gpt-5-mini for extraction and classification; nano input is 25 times cheaper than gpt-5.
- Try Flex for interactive but non-urgent calls and keep a fallback to standard processing.
- Set a hard spend limit and a spend alert on the organization before you ship a key anywhere.
The short version
- No free trial credit appears on OpenAI’s pricing or rate limits pages; the Free tier allows $100 of usage per month in approved geographies.
- Credits are prepaid, and $5, $100 and $500 of purchases move you to the Build, Launch and Grow tiers automatically.
- Batch and Flex both halve the GPT-5 family’s token prices; cached input costs 10% of standard input.
- Spend alerts notify, hard spend limits stop traffic with a 429, and enforcement lags slightly.
- Startup credit amounts come through partners, and OpenAI publishes no figure of its own that we can quote.
FAQ
Does openai api give free credits
No, not on any official page. New organizations in approved geographies sit in a Free usage tier with a $100 monthly cap, and the first paid tier requires $5 of credit purchases.
How do openai usage tiers work now
There are four tiers: Free, Build, Launch and Grow. Build, Launch and Grow require $5, $100 and $500 in total credit purchases and allow $500, $5,000 and $200,000 of monthly usage. Upgrades happen automatically with no waiting period listed.
How much does the openai batch api save
50%. gpt-5 drops from $1.25 to $0.625 per 1M input tokens and from $10.00 to $5.00 per 1M output tokens, with results returned within 24 hours.
What is openai cached input pricing
Cached input tokens on the GPT-5 family cost 10% of the standard input rate, for example $0.125 instead of $1.25 per 1M on gpt-5. Caching is automatic once a prompt prefix reaches 1,024 tokens on current models.
Does chatgpt plus include api credits
No. ChatGPT Plus is a consumer subscription for the chat product. The API is billed separately through prepaid credits on the platform billing page.