🇺🇸 USA · OpenAI API

Status: 🟩 COMPLETE 🟦 LIVING Last updated: 2026-06-26 Plain-English tagline: OpenAI’s developer platform — direct API access to GPT, gpt-image-1, Sora, Whisper, Voice Engine, Codex CLI, embeddings, and all OpenAI’s models. Distinct from ChatGPT (consumer) and Azure OpenAI Service (enterprise via Microsoft).


Front-matter facts

FieldValue
VendorOpenAI (San Francisco, USA)
Country / origin🇺🇸 USA
Recommended for Australian users?✅ Yes — fully accessible from AUS; AUD card accepted; pre-paid credits or post-paid invoicing
Privacy summaryNo training on API inputs by default (since March 2023); inputs retained 30 days for abuse-monitoring then deleted; Limited Access exemption available for compliance
Free tierFree trial credit on sign-up (varies — sometimes US$5-18); free-tier rate limits on some models
Paid tiersPay-per-token; pre-paid credits; AUS card accepted; Enterprise via direct sales
First releasedJune 2020 (GPT-3 API)
Last reviewed2026-06-26
Official sitehttps://platform.openai.com

What it is

The OpenAI API is OpenAI’s developer platform at platform.openai.com. It’s the direct way to use OpenAI’s models in your own code, separate from:

  • ChatGPT at chatgpt.com (consumer chat product)
  • Azure OpenAI Service (enterprise distribution via Microsoft)
  • DuckDuckGo / Brave Leo / Kagi (indirect access via privacy intermediaries)

API capabilities:

  • Chat completions — GPT-5, GPT-5 mini, GPT-5 nano, o-series reasoning, GPT-4o legacy
  • Image generation — gpt-image-1, DALL-E 3 (legacy)
  • Video generation — Sora
  • Voice — TTS API, Voice Engine (limited access), Whisper (speech-to-text)
  • Embeddings — text-embedding-3-small / large
  • Realtime API — bi-directional voice streaming (real-time agents)
  • Codex — coding agent CLI (open-source, free; uses API)
  • Operator / ChatGPT Agent — browser-based agent (limited API surface)
  • Fine-tuning — adapt models on your data
  • Batch API — async processing at 50% discount
  • Assistants API — stateful agent orchestration
  • Vector stores / Files — for RAG workflows
  • Function calling / Tool use — agent capabilities

What you’d use it for

  • Build apps that call GPT models in your own code
  • Embed AI features in your products
  • Custom chatbots with your own UI / branding
  • Image generation via API for your apps
  • Voice / transcription for your products
  • Agent workflows — Assistants API, Realtime API
  • Bulk processing — Batch API for cheap async work
  • Codex CLI for terminal-based coding agent
  • Fine-tuning GPT on your data for specialised use

When NOT to use OpenAI API:

  • Personal chat use (use ChatGPT consumer instead — usually cheaper for the same use)
  • Enterprise needing Azure compliance (use Azure OpenAI Service)
  • Best-of-breed multi-model strategy (use AWS Bedrock / Vertex AI / OpenRouter)
  • AUS data residency required (use Azure OpenAI Service in Australia East)

How to sign up + first 5 minutes from Australia

  1. Go to platform.openai.com. Sign up with email / Google / Apple / Microsoft account.
  2. Note: this is SEPARATE from ChatGPT — even if you have ChatGPT Plus, you need to sign up here too.
  3. Verify email + phone (Australian mobile (+61) accepted)
  4. Add payment method:
    • AUS Visa / Mastercard / Amex accepted
    • Pre-paid credits model — load credits, API draws from them
    • Minimum top-up US$5
    • Auto-top-up optional
  5. Get an API key — Dashboard → API keys → Create new
  6. Try it:
    from openai import OpenAI
    client = OpenAI(api_key="sk-...")
    response = client.chat.completions.create(
        model="gpt-5",
        messages=[{"role": "user", "content": "Hello"}]
    )
  7. Set usage limits — Settings → Limits → set monthly cap to prevent surprise charges

What it costs

Pay-per-token (most use)

Examples (verify current at openai.com/api/pricing):

ModelInput ($ per million tokens)Output ($ per million tokens)
GPT-5~US$3~US$15
GPT-5 mini~US$0.15~US$0.60
GPT-5 nano~US$0.05~US$0.20
o-series reasoningHigher (~US$15-60)Higher
gpt-image-1Per-image (~US$0.02-0.08)N/A
SoraPer-second videoN/A
Whisper~US$0.006 / minute audioN/A
TTS~US$15 per million charactersN/A
text-embedding-3-small~US$0.02 per millionN/A

Batch API

  • 50% discount on async processing (24-hour turnaround)
  • Useful for bulk extraction / classification / generation

Provisioned Throughput

  • Reserved capacity for guaranteed latency at scale

Free trial credit

  • Often US$5-18 for new accounts (verify current)

How it compares to alternatives

AspectOpenAI API directAzure OpenAI ServiceAWS BedrockVertex AIAnthropic API direct
Model breadthOpenAI onlyOpenAI onlyMulti-vendorMulti-vendorAnthropic only
Cheapest for OpenAIYes (direct)Slight markupN/AN/AN/A
AUS data residencyLimitedYes (Australia East)Yes (Sydney)Yes (Sydney + Melbourne)Via Bedrock
Enterprise contractsDirect OpenAIMicrosoft enterpriseAWS enterpriseGoogle enterpriseAnthropic enterprise
Newest features firstYesSlight lagN/AN/AYes (for Claude)
Cost (token pricing)Lowest for OpenAISlightly higherSlightly higherSlightly higherLowest for Claude
Best forDirect OpenAI prototyping + productionMicrosoft-stack enterpriseMulti-vendor AWSMulti-vendor GCPDirect Anthropic

For most developers prototyping or building products with OpenAI specifically, direct OpenAI API is the simplest and cheapest path. For AUS enterprise with data-residency, Azure OpenAI Service in Australia East wins.


Privacy / data handling

  • No training on API inputs by default — contractually committed since March 2023
  • Inputs retained 30 days for abuse-monitoring, then deleted
  • Limited Access exemption available (no retention) for high-compliance customers
  • Enterprise customers get tenant-isolated + additional commitments
  • Tier 1+ accounts can opt out of zero-data-retention concerns via enterprise contract
  • Data flows to US data centres by default; no native AUS data residency for direct OpenAI API (use Azure OpenAI Service for that)

Recent changes

  • 2026: GPT-5 family broadly available via API; gpt-image-1 generally available; Sora API access; Codex CLI open-source
  • 2025: Codex agent products launched
  • 2024: GPT-4o and o-series; multimodal expanded
  • March 2023: No-training-by-default policy announced
  • June 2020: GPT-3 API original launch

Gotchas

  • Separate from ChatGPT — ChatGPT Plus US$20/mo does NOT include API credits
  • Pre-paid credits don’t roll over indefinitely — verify expiration policy
  • Set usage limits to prevent surprise spending
  • Rate limits scale with tier — new accounts have lower limits; build to higher tier through usage history
  • AUS billing uses USD with conversion + fees on most cards; use fee-free card (Wise / Revolut / ANZ Plus / UP / ING Orange One)
  • Latest model availability — new models sometimes have rolling region availability
  • Function calling format changed in late 2025 with structured outputs — verify your client library is current
  • Cost-optimisation matters — pick the right model tier (GPT-5 nano for simple tasks vs GPT-5 Pro for hard reasoning)
  • Realtime API is brilliant for voice but consumes audio tokens — model usage carefully

See also


Sources