๐Ÿ‡บ๐Ÿ‡ธ USA ยท OpenAI GPT models lineup

Status: ๐ŸŸฉ COMPLETE ๐ŸŸฆ LIVING Last updated: 2026-06-26 Plain-English tagline: OpenAIโ€™s family of models โ€” GPT-5, GPT-5 mini, GPT-5 nano, GPT-5 Pro, o-series reasoning models, Sora, gpt-image-1, Whisper. The full catalog explained.


Front-matter facts

FieldValue
VendorOpenAI (San Francisco, USA)
Country / origin๐Ÿ‡บ๐Ÿ‡ธ USA
Recommended for Australian users?โœ… Yes
Privacy summaryPer-product privacy posture; API never trains by default; consumer ChatGPT trains by default with opt-out
Free tierMany models available free via ChatGPT free or limited API trial credits
Paid tiersVia ChatGPT subscriptions, API pay-per-token, Azure OpenAI Service
First releasedGPT-1 2018; GPT-5 family 2025-26
Last reviewed2026-06-26
Official sitehttps://platform.openai.com/docs/models

What it is

OpenAI ships a family of AI models across several capabilities. This entry is the lineup explainer โ€” what each model is for, when to choose which.

Chat / general-purpose

ModelBest forApproximate cost (per M tokens, in/out)
GPT-5Default best-quality chat~US15
GPT-5 miniMost production use; balanced cost-perf~US0.60
GPT-5 nanoCheap classification / extraction at scale~US0.20
GPT-5 ProHardest reasoning; longest agent runsPremium pricing
GPT-4o (legacy)Still available; replaced by GPT-5 defaultLower than GPT-5
GPT-4o mini (legacy)Replaced by GPT-5 miniCheap

Reasoning (o-series)

ModelBest for
o3Hard math / science / coding reasoning
o4-miniCheaper reasoning
GPT-5 Pro with extended thinkingCombined reasoning + general capability

Image generation

ModelBest for
gpt-image-1 (โ€œImageโ€ / โ€œImage 2.0โ€)Current flagship image gen; in-chat editing
DALL-E 3 (legacy)Replaced by gpt-image-1; still accessible via API
DALL-E 2 (legacy)Deprecated

Video generation

ModelBest for
Sora 2 / Sora TurboCurrent video gen; with audio in Sora 2
Sora 1 (legacy)Replaced by Sora 2

Voice

ModelBest for
Voice EngineVoice cloning (limited access)
TTS (text-to-speech API)High-quality voice generation
WhisperSpeech-to-text transcription
Realtime APIBi-directional voice streaming

Embeddings

ModelBest for
text-embedding-3-largeHigh-quality embeddings for RAG
text-embedding-3-smallCheaper / faster embeddings

Coding

ModelBest for
codex-1 (the 2025 agent model)Powers OpenAI Codex agent CLI + cloud
GPT-5 familyGeneral coding via API / ChatGPT

Open-weight (rare for OpenAI)

ModelBest for
gpt-ossOpenAIโ€™s open-weight releases (Llama / Mistral alternative; smaller selection)

How to choose between them

QuestionAnswer
Best quality for occasional use?GPT-5 (default in ChatGPT)
Cheap at scale?GPT-5 mini or nano
Hardest reasoning / agent tasks?GPT-5 Pro or o-series
Image generation?gpt-image-1
Video generation?Sora 2
Transcription?Whisper
Voice synthesis?TTS or Voice Engine
Real-time voice?Realtime API
Embeddings for RAG?text-embedding-3-large

Pricing tips

  • Free: ChatGPT free tier (limited GPT-5)
  • US$20/month: ChatGPT Plus (Plus tier of all)
  • US$200/month: ChatGPT Pro (heavy use, Operator, Pulse)
  • API metered: pay-per-token (cheapest for batch / programmatic)
  • Batch API: 50% discount on async work
  • Azure OpenAI Service: enterprise distribution with AUS data residency

Recent changes

  • 2026: GPT-5 family broadly deployed; GPT-5 Pro
  • 2025: GPT-5 release; gpt-image-1 replaced DALL-E 3; Sora 2 launched
  • 2024: GPT-4o, o1 (later renamed o3 / o4-mini), Sora 1
  • 2023: GPT-4 launched
  • 2022: ChatGPT launched (GPT-3.5)
  • 2018: GPT-1 (foundational research)

Gotchas

  • Model versions matter โ€” always pin specific model (e.g., gpt-5-2025-08-07) in production rather than generic gpt-5 alias
  • o-series tokens are more expensive but produce reasoning tokens (which contribute to cost)
  • Sora video pricing is per-second; can add up fast
  • gpt-image-1 in ChatGPT is included with Plus / Pro; via API metered per image
  • Older models (GPT-4o, GPT-3.5) are deprecated for new use โ€” migrate

See also


Sources