HomeFull speed ahead. Keep track of advancements in every frontier

TheTechno Optimist.

CodexOpenAI—StargateOpenAI—OpenAI—GPT modelsOpenAI—OpenAI—gpt-ossOpenAI—CodexOpenAI—GPT modelsOpenAI—

OpenAI

ChatGPT, GPT models, Codex, Stargate

US$852B privateest. 2015openai.com (opens openai.com)Checked 25 Sep

The brief

Analyst view
  • Makes ChatGPT and the GPT-5.x models. GPT-5.6 came out in July 2026.
  • Raised $122B at an $852B valuation in March 2026. Amazon, NVIDIA and SoftBank were the biggest investors.
  • Lining up tens of gigawatts of computing power through Stargate and chip deals with NVIDIA, AMD and others.

Technical approach

As reported
Scale

Bigger is better

OpenAI's core bet: language models get predictably smarter with more size, data and compute. Its 2020 paper measured this.

Post-training

Learning from human feedback

People rate answers and the model learns to prefer the good ones. This turned GPT-3.5 into ChatGPT.

Reasoning

Thinking before answering

Since o1 (2024), models are trained to reason step by step first. More thinking time gives better answers.

Products

Routers and agents

GPT-5 sends easy questions to a fast model and hard ones to a reasoning model. Codex adds tools to finish tasks.

Programs

OpenAI, four programs

Updated 9 Jul 2026
GPT modelsOpenAI's main AI models, used in ChatGPT and sold to developers. Since GPT-5, one system picks a fast or a deep-thinking model for each question.ScalingNextAI research intern-level system (stated goal) · Sep 2026Updated 9 Jul 2026View program →CodexAn agent that reads a codebase, writes and tests changes in a safe sandbox and proposes them for review. The name was first used in 2021 for the model behind GitHub Copilot.ScalingUpdated 5 Feb 2026View program →StargateA joint venture of OpenAI, SoftBank, Oracle and MGX to spend up to $500B on about 10 GW of US AI data centers. Sites in the UAE, Norway and elsewhere followed.DeployNextFull ~10 GW US buildout · 2029Updated 30 Oct 2025View program →gpt-ossOpenAI's first open-weight models since GPT-2. The larger one runs on a single high-end GPU, the smaller one on a laptop.DeployUpdated 29 Oct 2025View program →
Research · Proto · Pilot · Deploy · Scaling

Research & demos

5 items
  • Sep 2025
    paperWhy Language Models Hallucinate

    Argued models make things up because training and tests reward confident guessing.

  • Sep 2024
    blogLearning to Reason with LLMsGPT models

    Introduced o1: answers got better with more training and more thinking time.

  • Mar 2023
    paperGPT-4 Technical ReportGPT models

    Showed a big multimodal model's progress could be predicted, and that it passed hard exams.

  • Mar 2022
    paperTraining language models to follow instructions with human feedback (InstructGPT)GPT models

    Set the recipe for turning a raw model into a helpful assistant using human ratings.

  • Jan 2020
    paperScaling Laws for Neural Language Models

    Showed models improve smoothly and predictably with more size, data and compute. This started the scaling era.

Open technical problems

Open · Active · Partial
  • Making things upActive research

    Models still state false facts confidently. OpenAI argues tests reward guessing over saying "I don't know".

  • Compute, power and costOpen

    Each new generation needs far more compute. Power, chips and money now limit progress as much as ideas.

  • Watching autonomous agentsActive research

    As agents run longer tasks alone, spotting cheating or misuse gets harder. OpenAI studies reading the model's reasoning.

Peers

Frontier models · 5

Company & funding

5 rounds · as reported
Valuation
$852Bprivate
Raised
$179Bdisclosed private rounds
Last round
Venture roundMar 2026
HQ
US
Founded
2015
$10B$100B$1T20232024202520262027$157B$300B$500B$852B

Post-money valuation, log scale

DateRoundAmountPost-moneyLead
Mar 2026Venture round$122B$852BAmazon, NVIDIA, SoftBank
Oct 2025Employee secondary sale$6.6B$500B—
Mar 2025Venture round$40B$300BSoftBank
Oct 2024Venture round$6.6B$157BThrive Capital
Jan 2023Microsoft multi-year investment$10B—Microsoft

Partners & customers

3 investors · 5 suppliers · 1 partner
  • Investor
    Microsoft

    Largest early backer; holds roughly 27% of OpenAI Group PBC after the October 2025 recapitalisation.

  • Investor
    SoftBank

    Led the 2025 $40B round and is a Stargate partner.

  • Investor
    Amazon

    Committed $50B in the March 2026 round; also an AWS compute supplier.

  • Supplier
    Oracle

    Builds and operates Stargate data centers, starting with Abilene, Texas.

  • Supplier
    NVIDIA

    Letter of intent (Sep 2025) to deploy at least 10 GW of NVIDIA systems.

  • Supplier
    AMD

    6 GW Instinct GPU agreement (Oct 2025); first 1 GW of MI450 from H2 2026.

  • Partner
    Broadcom

    Co-developing custom OpenAI accelerators, 10 GW announced Oct 2025.

  • Supplier
    CoreWeave

    GPU cloud contracts worth over $20B.

  • Supplier
    Cerebras

    Multi-year deal (Jan 2026) for wafer-scale inference capacity.

People

Key people
  • Sam AltmanCEO and co-founder
  • Greg BrockmanPresident and co-founder
  • Jakub PachockiChief Scientist
Primer