Skip to content

OpenAI

Dossier checked 28 Sep 2026

ChatGPT on the free plan GPT-6.1 Sol, GPT-6 Sol and GPT-6 Astra are not on the free plan. GPT-6.1 Terra is not named for free users. The Access Gap, checked 3 Oct 2026.

What it sells now

Newest frontier release

GPT-6.1 Terra

26 Oct 2026 · the lab’s announcement

Fast, efficient model for visual reasoning, computer use and writing, released in the API; rollout to ChatGPT Work and Codex for Plus and higher plans began the same day

Flagship list price, per million tokens

$10 in $50 out

GPT-6 Astra, OpenAI’s most capable model (prompts up to 272K tokens), released 4 Sep 2026 · pricing page

GPT-6 Sol and Luna came later in September as cheaper tiers of the same family; GPT-6.1 Sol followed on 29 September, and GPT-6.1 Terra on 26 October at $1.00 in and $5.00 out.

Weights

Closed flagship, some open

Partial. gpt-oss-120b and gpt-oss-20b, and the gpt-oss-safeguard pair, are open under Apache 2.0; every GPT-5 and GPT-6 model is closed. Source

Where you can buy the flagship

Source

13 releases in the year to 30 Oct 2026, 6 the year before · 29 in all since 2018

Our read, not a fact

In the model

The question we ask of every lab: is most of its value in the model, which gets cheaper fastest, or in something other work is built on? Our answer: in the model itself.

OpenAI’s scarcity sits on the output layer, the part of the stack that commoditises fastest. Its durable bet is ChatGPT as a daily habit and the platform that routes work to whichever model wins. The frontier model itself is the commodity.

What would change our readChatGPT loses its consumer default to a rival with better distribution, or frontier quality stops being matchable by the next lab within months.

01 · What it runs on, and the money

Trains and serves on

NVIDIA GPUs carry the training fleet and most of the inference stack, by OpenAI’s own account. Around them sit clouds from Microsoft, Oracle, AWS, CoreWeave and Google Cloud, and chips from AMD, AWS Trainium, Cerebras and a design co-developed with Broadcom. Source

What it depends on Five clouds and five chip suppliers, yet OpenAI says NVIDIA still carries training and most inference, so one chip vendor sits under most of its compute. Source

Biggest risk NVIDIA GPUs carry the training fleet and most of the inference stack, which makes NVIDIA supply its largest physical dependency. Source

Raised
$122B committed in the March 2026 round, anchored by Amazon, NVIDIA and SoftBank Source
Valuation
$852B post-money (31 Mar 2026) Source
Revenue
$2B a month, about $24B annualised (31 Mar 2026), the last figure OpenAI has confirmed Source
Structure
The nonprofit OpenAI Foundation controls OpenAI Group PBC, a Delaware public benefit corporation, through special voting rights that let it appoint and replace any director. Source

NotesThe flagship list price has gone up: GPT-6 Astra lists at eight times GPT-5’s input price and five times its output price, while the cheapest GPT-6 tier, Luna, lists at $0.10 in and $0.50 out. Press reports put revenue above $40B a year (Bloomberg, Aug 2026) and talks for a round above $1.2T (Bloomberg, Sep 2026); OpenAI has confirmed neither. Headcount and total funding are not published.

02 · Releases

data through 30 Oct 2026

  1. OpenAI 23 GPT-6.1 Terra26 Oct 2026

Each dot is a frontier release at its date. Click one to find it in the full log. Every lab on one chart · Which plans and apps its new models reached

The landmarks

  1. 4 Sep 2026 GPT-6 Astra First GPT-6 model; wide availability a day after a limited start First GPT-6
  2. 9 Jul 2026 GPT-5.6 (Sol / Terra / Luna) Three price-performance tiers, GA across ChatGPT, Codex and the API FlagshipTiered
  3. Apr 2026 GPT-5.5 Flagship system of the 5.5 line Flagship
  4. Aug 2025 GPT-5 Unified system routing across model tiers UnifiedFlagship
  5. Sep 2024 o1-preview First inference-time reasoning model ReasoningFirst of kind
  6. May 2024 GPT-4o Omni real-time text, vision and audio OmniMultimodal
  7. Mar 2023 GPT-4 Multimodal flagship, large capability jump MultimodalFlagship
  8. Nov 2022 ChatGPT Consumer chatbot; the fastest app to reach 100M users Consumer launch
  9. Jun 2020 GPT-3 175B-parameter model behind a public API 175BAPI
  10. Jun 2018 GPT-1 First generative pretrained transformer First GPTResearch
The full release logall 29, newest first
  1. 2026
  2. 26 Oct 2026 GPT-6.1 Terra Frontier Fast, efficient model for visual reasoning, computer use and writing, released in the API; rollout to ChatGPT Work and Codex for Plus and higher plans began the same day Source
  3. 29 Sep 2026 GPT-6.1 Sol Frontier Lower-cost professional and agentic model released in the API; rollout to ChatGPT Work and Codex began the same day Source
  4. 22 Sep 2026 GPT-6 Sol / Luna Frontier Two GPT-6 models released together Source
  5. 4 Sep 2026 GPT-6 Astra Frontier First GPT-6 model; wide availability a day after a limited start Source
  6. 9 Jul 2026 GPT-5.6 (Sol / Terra / Luna) Frontier Three price-performance tiers, GA across ChatGPT, Codex and the API Source
  7. 5 May 2026 GPT-5.5 Instant Frontier ChatGPT’s new default for everyone, the tier just below GPT-5.5 Thinking; on the API as chat-latest Source
  8. Apr 2026 GPT-5.5 Frontier Flagship system of the 5.5 line
  9. Mar 2026 GPT-5.4 Frontier Point release
  10. 17 Mar 2026 GPT-5.4 mini GPT-5.4’s mini, at $0.75 in and $4.50 out, released to the API with GPT-5.4 nano; third that day, below GPT-5.4 and GPT-5.3 Instant Source
  11. 3 Mar 2026 GPT-5.3 Instant Frontier ChatGPT’s new default for everyone, the tier just below Thinking; on the API as gpt-5.3-chat-latest Source
  12. Feb 2026 GPT-5.3-Codex Frontier Coding-focused Codex variant
  13. 2025
  14. Dec 2025 GPT-5.2 Frontier Point release
  15. Nov 2025 GPT-5.1 Frontier Iterative quality and steerability upgrade
  16. Aug 2025 GPT-5 Frontier Unified system routing across model tiers
  17. Apr 2025 o3 / o4-mini Frontier Full reasoning lineup with tool use
  18. Apr 2025 GPT-4.1 Frontier API-first coding and instruction model
  19. Feb 2025 GPT-4.5 Frontier "Orion": a large non-reasoning research preview
  20. Jan 2025 o3-mini Efficient reasoning at lower cost
  21. 2024
  22. Dec 2024 o1 Frontier Full release of the o1 reasoning model
  23. Sep 2024 o1-preview Frontier First inference-time reasoning model
  24. May 2024 GPT-4o Frontier Omni real-time text, vision and audio
  25. 2023
  26. Nov 2023 GPT-4 Turbo Frontier Cheaper, faster, longer context
  27. Mar 2023 GPT-4 Frontier Multimodal flagship, large capability jump
  28. 2022
  29. Nov 2022 text-davinci-003 Frontier OpenAI’s most capable model on its API, released two days before ChatGPT
  30. Nov 2022 ChatGPT Frontier Consumer chatbot; the fastest app to reach 100M users
  31. Mar 2022 GPT-3.5 InstructGPT-tuned for following instructions
  32. 2020
  33. Jun 2020 GPT-3 175B-parameter model behind a public API
  34. 2019
  35. Feb 2019 GPT-2 Staged release over 2019 on safety concerns
  36. 2018
  37. Jun 2018 GPT-1 First generative pretrained transformer

03 · When to reach for them

practical read

Broadest reach
Reach for it when you want the widest consumer and developer footprint (ChatGPT, Nov 2022).
Hardest end-to-end work
Pick GPT-6 Astra for long coding, computer-use and research tasks (Sep 2026).
Price tiers in one family
Use GPT-6 Sol for most agent work and GPT-6 Luna for high-volume tasks, at a fraction of Astra’s price (Sep 2026).

04 · Weights, across all ten

the flagship

Open flagship
DeepSeek Mistral AI Alibaba (Qwen) Moonshot AI Zhipu (Z.ai)
Closed flagship, some open
OpenAI Google Meta xAI
Closed
Anthropic

05 · Essays that read OpenAI

from the archive

The Model Is Not the Moat If frontier capability keeps centralising, the durable edge shifts outward into trust, workflow fit, and the surrounding package. You're Not Comparing Models. You're Comparing Contracts. Agent benchmarks don't measure models. They measure contracts. Two teams running the same model can publish different scores, and both can be honest. The Forest Floor Is The Product Most people are optimising for canopy. The work that survives the next five AI releases is built from the forest floor up.

06 · About OpenAI

at a glance

Founded
2015
HQ
San Francisco, USA
Leadership
Sam Altman (CEO)
Backers
Microsoft, SoftBank, NVIDIA, Amazon
Also builds
ChatGPT, Codex, API platform
Known for
ChatGPT, GPT series

Founded in December 2015 by Sam Altman, Greg Brockman, Ilya Sutskever, Elon Musk and others as a nonprofit lab. It set up a capped-profit arm in 2019 and in October 2025 became OpenAI Group PBC, controlled by the nonprofit OpenAI Foundation. Headquartered in San Francisco, it is best known for the GPT series and ChatGPT.