OpenAI

OpenAI stands as the most influential artificial intelligence company in history, transforming from a $130 million nonprofit in 2015 into a $500 billion behemoth by late 2025. Founded to ensure AI benefits humanity, the company has become synonymous with both breakthrough technology and fierce controversy, from launching ChatGPT to weathering a CEO firing that nearly destroyed it.

This guide documents OpenAI’s complete journey: the founding vision, the Microsoft billions, every major product launch, the dramatic boardroom coup, the safety researcher exodus, and the company’s current dominance amid intensifying competition.

Quick facts

The founding and early years (2015-2019)

December 11, 2015: OpenAI arrives

OpenAI was announced on December 11, 2015, as a nonprofit artificial intelligence research company. The organization was incorporated in Delaware and initially operated from Greg Brockman’s San Francisco apartment before moving to the Pioneer Building in the Mission District.

The co-chairs:

• Sam Altman — President of Y Combinator
• Elon Musk — CEO of Tesla and SpaceX

The founding leadership:

• Greg Brockman — CTO (former CTO of Stripe)
• Ilya Sutskever — Research Director (recruited from Google, studied under Geoffrey Hinton)

The founding research team: Trevor Blackwell, Vicki Cheung, Andrej Karpathy, Durk Kingma, John Schulman, Pamela Vagata, and Wojciech Zaremba.

The original mission

The founding announcement stated OpenAI’s purpose: “To advance digital intelligence in the way that is most likely to benefit humanity as a whole, unconstrained by a need to generate financial return.”

The nonprofit structure would allow OpenAI to “build value for everyone rather than shareholders.” Researchers were “strongly encouraged to publish their work” and any patents would “be shared with the world.”

Why OpenAI was created

Concerns about Google’s dominance: Google had acquired DeepMind in 2014, which had declared pursuit of AGI. Musk reportedly had a “bitter argument” with Google co-founder Larry Page about AI in summer 2015, Page saw AI as humanity-elevating while Musk warned of existential doom. They reportedly stopped speaking afterward.

AI safety fears: Musk had publicly called AI “potentially more dangerous than nukes” in 2014. The founders believed AGI development needed a safety-focused counterweight to commercial labs.

Democratisation concerns: Powerful AI was concentrating in wealthy corporations. The “Open” in OpenAI reflected commitment to openness and transparency.

The funding reality versus the headlines

While OpenAI announced a $1 billion pledge, the reality proved different. According to MIT Technology Review’s analysis of IRS filings, only approximately $130-133 million was actually donated to the nonprofit from inception through 2019.

Key donors:

• Elon Musk — OpenAI states “less than $45M”; Musk claimed $100M (later revised to $50M)
• Sam Altman — ~$3.78 million (loaned then forgave with interest)
• Reid Hoffman — ~$6 million via Aphorism Foundation
• Open Philanthropy (Dustin Moskovitz/Cari Tuna) — $30 million across 2017-2019
• Amazon Web Services — Cloud computing services

Musk’s departure

On February 20, 2018, Elon Musk left OpenAI’s board. The official reason cited potential conflicts with Tesla’s growing AI focus for autonomous driving.

The behind-the-scenes story proved more complex. According to OpenAI’s own account, in early 2018, Musk told Altman he believed OpenAI had “fallen fatally behind Google.” He proposed taking control and running it himself, wanting “majority equity, initial board control, and to be CEO.” When the founders rejected this, Musk suggested merging OpenAI into Tesla. This too was rejected.

Musk departed stating OpenAI’s “probability of success was 0” and stopped funding the organisation. Reid Hoffman stepped in to bridge the funding gap.

Early research milestones

2016: OpenAI Gym released (reinforcement learning platform); Universe platform unveiled

August 2017: OpenAI bot defeated professional Dota 2 player Dendi in 1v1 matches

June 11, 2018: GPT-1 released — 117 million parameters, trained on BooksCorpus. This breakthrough demonstrated pre-trained transformers could improve NLP tasks.

April 13, 2019: OpenAI Five defeated OG (The International 2018 world champions) 2-0, the first AI to beat world champions in esports.

The capped-profit pivot

On March 11, 2019, OpenAI announced creation of “OpenAI LP,” a radical new structure. The compute costs were crushing, doubling every 3.4 months, and the original $1B pledge had yielded only ~$130M. Leadership realised in 2017 that AGI would require “billions of dollars per year.”

How capped-profit works:

• Investors receive returns capped at a multiple of their investment
• First round cap: 100x (e.g., $10M investment = maximum $1B return)
• OpenAI Nonprofit remains the controlling General Partner
• Any excess returns beyond caps go to the nonprofit

This structure enabled the Microsoft investment that would transform OpenAI.

The Microsoft partnership

Investment timeline

July 22, 2019 — $1 billion: Microsoft announced its first major investment, split approximately half in cash and half in Azure credits. Microsoft became OpenAI’s exclusive cloud provider and preferred partner for commercialising AI technologies.

2021 — ~$2 billion: Microsoft made additional investments (exact amount not officially disclosed). Microsoft built a top-5 supercomputer with 285,000+ CPU cores and 10,000 GPUs for OpenAI.

January 23, 2023 — $10 billion: Following ChatGPT’s explosive success, Microsoft invested approximately $10 billion over multiple years, bringing total cumulative investment to ~$13 billion.

October 2024 — $750 million: Microsoft participated in OpenAI’s $6.6B round at $157 billion valuation.

Ownership and economic arrangements

Pre-2025: Microsoft did not own traditional equity but was entitled to up to 49% of profits from OpenAI’s capped-profit subsidiary.

October 28, 2025 restructure: Microsoft now holds 27% equity in OpenAI Group PBC, valued at approximately $135 billion. The full ownership breakdown:

• Microsoft: 27%
• OpenAI Foundation (nonprofit): 26%
• Employees and other investors: 47%

The Azure exclusivity deal

Azure serves as OpenAI’s exclusive cloud provider for all API workloads. The October 2025 restructure extended Microsoft’s IP rights through 2032, including post-AGI models. OpenAI has committed to purchasing $250 billion in Azure services.

However, OpenAI gained flexibility for non-API infrastructure, enabling the Stargate Project, a $500 billion joint venture with SoftBank and Oracle announced January 2025.

Revenue sharing mechanics

The relationship involves bidirectional revenue sharing:

• OpenAI pays Microsoft 20% of its revenue
• Microsoft pays OpenAI 20% of revenue from Azure OpenAI Service and Bing
• In 2024, Microsoft received $493.8 million in revenue share from OpenAI
• Through Q3 2025: $865.8 million

Complete product launch timeline

2019: GPT-2 and the “too dangerous” controversy

February 14, 2019: OpenAI announced GPT-2 but initially withheld the full model, citing concerns about potential misuse for generating fake news and disinformation at scale. The 1.5 billion parameter model was trained on 40GB of internet text.

The staged release sparked fierce debate. Critics accused OpenAI of exaggerating risks for media attention. After monitoring showed “no strong evidence of misuse,” the full model was released November 5, 2019.

2020: GPT-3 changes everything

June 11, 2020: GPT-3 API launched in private beta. At 175 billion parameters, it was 10x larger than any previous non-sparse model.

API pricing at launch:

GPT-3’s “few-shot learning” capability, performing tasks with minimal examples, demonstrated a fundamental breakthrough in AI capabilities.

2021-2022: Image generation and code

January 5, 2021: Original DALL-E revealed as a research demonstration, a 12-billion parameter GPT-3 variant modified for image generation.

August 10, 2021: Codex released, powering GitHub Copilot (launched June 2021 in technical preview, generally available June 2022 at $10/month).

April 6, 2022: DALL-E 2 announced with 4x greater resolution and diffusion-based architecture. Waitlist removed September 28, 2022. Pricing: 50 free credits first month, then $15 for 115 credits.

September 21, 2022: Whisper open-sourced, a speech recognition model trained on 680,000 hours of multilingual data, supporting ~99 languages.

November 30, 2022: ChatGPT launches

ChatGPT launched as a free research preview with minimal fanfare, just a blog post. Based on GPT-3.5 fine-tuned with Reinforcement Learning from Human Feedback (RLHF), it became the fastest-growing consumer application in history:

• 1 million users: 5 days
• 57 million monthly users: First month
• 100 million monthly users: ~2 months

The launch triggered “code red” at Google, which rushed Bard to market. Microsoft announced $10B in new investment and integrated ChatGPT into Bing within months.

2023: GPT-4 and the enterprise push

March 14, 2023: GPT-4 released, a large multimodal model accepting both image and text inputs. It scored in the 90th percentile on the Bar Exam versus GPT-3.5’s 10th percentile.

September 20, 2023: DALL-E 3 announced with native ChatGPT integration and dramatically improved text rendering.

November 6, 2023 (DevDay):

• GPT-4 Turbo: 128K context window, 3x cheaper than GPT-4
• GPTs and GPT Store announced (store launched January 10, 2024)
• Assistants API for agent-like experiences
• Text-to-Speech models released

2024: The multimodal year

May 13, 2024: GPT-4o (“omni”) released, natively multimodal with text, audio, image, and video inputs. Average response time: 320ms (human-like). 2x faster and half the price of GPT-4 Turbo.

July 18, 2024: GPT-4o mini replaced GPT-3.5 Turbo, 60% cheaper with an 82% MMLU score versus 69.8%.

September 12, 2024: o1 reasoning models previewed (codenamed “Strawberry,” formerly “Q*”). These models “think” before answering using chain-of-thought reasoning, excelling at complex reasoning, science, and math.

December 9, 2024: Sora publicly released during “12 Days of OpenAI”, generating 10-20 second videos up to 1080p resolution.

December 20, 2024: o3 preview announced, scoring 87.5% on ARC-AGI (versus o1’s 25-32%).

2025: GPT-5 arrives

January 31, 2025: o3-mini released, 63% cheaper than o1-mini with three thinking variants (low/medium/high).

February 27, 2025: GPT-4.5 “Orion” released as research preview, the last non-chain-of-thought model, focused on “EQ” and reduced hallucinations. Pricing: $75/1M input tokens (very expensive); deprecated by August 2025.

April 14, 2025: GPT-4.1 family released, GPT-4.1, GPT-4.1 mini, and GPT-4.1 nano with 1 million token context windows.

April 16, 2025: o3 and o4-mini released, o3 achieving 69.1% on SWE-bench Verified, o4-mini achieving 92.7% on AIME 2025.

August 7, 2025: GPT-5 released, a unified model combining reasoning (o-series) and fast responses (GPT-series). A real-time router automatically selects appropriate mode. Available to all users including free tier. Performance: 94.6% on AIME 2025, ~45% fewer hallucinations than GPT-4o.

November 13, 2025: GPT-5.1 released with dynamic thinking time adaptation and “no reasoning” mode option. GPT-5.1-Codex-Max added for agentic coding.

The Sam Altman firing crisis

Five days that shook Silicon Valley

Friday, November 17, 2023, ~noon PST: Sam Altman was informed of his removal 5-10 minutes before it happened via Google Meet. The board publicly announced his ouster, stating he was “not consistently candid in his communications.” Microsoft learned about it one minute before the announcement.

Mira Murati was named interim CEO. That evening, Greg Brockman quit in solidarity, followed by three senior researchers.

Saturday, November 18: The board agreed in principle to resign and reinstate Altman but missed their self-imposed 5 PM deadline. Investors including Tiger Global, Sequoia, and Microsoft pushed for Altman’s return.

Sunday, November 19: Altman and Brockman appeared at headquarters to negotiate. By evening, negotiations collapsed when the board chose Emmett Shear (former Twitch CEO) as new CEO. At midnight, Satya Nadella announced Microsoft was hiring Altman to lead a new AI research team.

Monday, November 20: 745 of 770 employees (~97%) signed a letter demanding board resignation, threatening to follow Altman to Microsoft. Remarkably, Ilya Sutskever, who had voted to fire Altman, signed the letter and posted: “I deeply regret my participation in the board’s actions.”

Tuesday, November 21: OpenAI announced “agreement in principle” for Altman’s return with a new board: Bret Taylor (Chair), Larry Summers, and Adam D’Angelo (lone holdover).

What the board actually alleged

Helen Toner, a board member who voted to fire Altman, later revealed specific allegations:

• The board “learned about ChatGPT on Twitter”, not informed before the November 2022 launch
• Altman didn’t disclose he personally owned the OpenAI Startup Fund despite being an “independent” board member
• He gave “inaccurate information about safety processes on multiple occasions”
• In October 2023, two executives approached the board with “screenshots and documentation” describing “psychological abuse” and evidence of “lying and being manipulative”

The speculated “Q*” breakthroughan AI reportedly solving grade-school math problems, was described as “a catalyst” but not the primary driver.

The aftermath

A WilmerHale investigation reviewed 30,000+ documents and concluded Altman’s conduct “did not mandate removal.” The prior board “acted within its broad discretion” but made errors, including not anticipating destabilisation.

Microsoft gained a non-voting board observer seat (later dropped in July 2024). By March 2024, the board expanded to include Sue Desmond-Hellmann, Nicole Seligman, Fidji Simo, and Altman himself rejoined.

Leadership and key departures

Current leadership (late 2025)

Notable: The CTO role remains unfilled since Mira Murati’s departure in September 2024.

The 2024 exodus

May 14, 2024 — Ilya Sutskever: The co-founder and chief scientist departed following months of tension after the November crisis. He founded Safe Superintelligence Inc. (SSI), which raised $1B+ and is valued at $32 billion as of late 2025.

May 15-17, 2024 — Jan Leike: The Head of Alignment publicly stated: “Safety culture and processes have taken a backseat to shiny products.” He joined Anthropic. The Superalignment team was disbanded following his departure.

August 2024 — John Schulman: The co-founder and head of alignment science left for Anthropic, later joining Mira Murati’s new company.

September 25, 2024 — Mira Murati, Bob McGrew, Barret Zoph: The CTO, Chief Research Officer, and VP Research all departed the same day. Murati founded Thinking Machines Lab, which raised $2B at $12B valuation in July 2025.

Remaining founders

Of the original 11 founding team members, only three remain at OpenAI:

• Sam Altman (CEO)
• Greg Brockman (President)
• Wojciech Zaremba (Lead of Language and Code Generation)

Business and financials

Revenue trajectory

The $2B ARR figure, often cited as “2023 revenue”, was actually reported by the Financial Times in February 2024 referring to December 2023 run rates.

Valuation history

The March 2025 round was the largest private tech funding round ever, $40B led by SoftBank’s $30B commitment, surpassing Ant Group’s $14B in 2018.

Subscription pricing

User metrics

ChatGPT’s weekly active users have grown 8x since November 2023:

• November 2023: 100 million
• August 2024: 200 million
• March 2025: 500 million
• October 2025: 800 million

92% of Fortune 500 companies use OpenAI products. Enterprise seats grew 9x year-over-year.

The cost challenge

OpenAI is not profitable. In 2024, the company lost approximately $5 billion against $3.7B revenue. Costs include:

• Running models (inference): ~$2B
• Training models: ~$3B
• Salaries: ~$700M

Projected losses through 2028 total $44 billion. The company targets cash-flow positive by 2029-2030.

Competition and market position

ChatGPT dominates but faces erosion

ChatGPT holds 61.3% market share as of December 2025 (75.4% including Microsoft Copilot), down from 76% in January 2024.

Anthropic: The enterprise challenger

Anthropic, founded in 2021 by former OpenAI executives Dario and Daniela Amodei, has emerged as the strongest challenger. Despite only 5% of consumer users, Anthropic generates 40% of OpenAI’s revenue due to superior enterprise monetisation: $211/monthly user versus OpenAI’s $25/weekly user.

Claude models lead in coding tasks. Microsoft now offers Anthropic models in GitHub Copilot and Azure, hedging its OpenAI bet.

Google Gemini catching up

Gemini 2.5 Pro ranks #1 on LMArena benchmarks and offers 1M-2M token context windows versus GPT-4’s 128K. Google’s integration advantages across Workspace, Search, and Android give it unique distribution. Gemini traffic grew 46% recently, faster than ChatGPT’s 7% quarterly growth.

Meta’s open-source disruption

Meta’s Llama models, released open-source with weights available, create competitive pressure by enabling free local deployment. DeepSeek, a Chinese model built on open-source principles, demonstrated that “open source can achieve state-of-the-art performance.” Meta doesn’t need to monetise AI models directly; benefits come from platform improvement.

xAI and others

Elon Musk’s xAI launched Grok 3, claiming benchmark superiority over GPT-4o (93.3% vs 79% on Math AIME’25). With a $113B valuation after merging with X, Grok offers API costs ~50% cheaper than GPT-4o and X platform integration.

Mistral AI (French, €450M+ funding) pursues open-source approaches competitive with GPT-3.5/4. Amazon Bedrock offers ~100 models including Anthropic and Meta, positioning AWS as a neutral platform.

Controversies and criticism

Safety researcher exodus

Throughout 2024, roughly half of OpenAI’s AI safety researchers left the company. The Superalignment team, announced July 2023 with promises of 20% of computing power, was disbanded May 2024.

Jan Leike’s departure statement captured the tension: “Building smarter-than-human machines is an inherently dangerous endeavour… But over the past years, safety culture and processes have taken a backseat to shiny products.”

In April 2025, OpenAI announced it would stop pre-release testing for persuasion/manipulation risks and consider releasing “high risk” models if competitors released similar, widely criticised as a “race to the bottom.”

Copyright litigation

The New York Times lawsuit (filed December 2023) alleges OpenAI scraped millions of articles without permission, seeking billions in damages and destruction of models trained on NYT content. In March 2025, Judge Sidney Stein rejected OpenAI’s motion to dismiss. OpenAI was ordered to produce 20 million ChatGPT chat logs as evidence. Trial date not yet set.

Authors Guild (filed September 2023): 17 authors including John Grisham and George R.R. Martin allege copyright infringement. Consolidated with the NYT case in October 2024.

GEMA (German music rights): OpenAI lost in Munich Regional Court in November 2024, found to have violated German copyright laws.

The Scarlett Johansson voice controversy

In September 2023, Altman approached Johansson to voice ChatGPT’s assistant, claiming her voice would be “comforting.” She declined. When GPT-4o launched in May 2024 with a voice called “Sky,” Johansson stated she was “shocked, angered and in disbelief” at the resemblance. Altman had tweeted the single word “her”, referencing the 2013 film where Johansson voices an AI.

OpenAI pulled the Sky voice and claimed it was recorded before Johansson outreach, though no lawsuit was filed.

Employee treatment and NDAs

In May 2024, Vox revealed that departing employees faced “extremely restrictive” NDAs with lifetime non-disparagement clauses. Violation could result in loss of all vested equity, potentially millions of dollars. Employees sometimes received just one week to sign.

Altman said he was “genuinely embarrassed” and “did not know this was happening.” OpenAI released former employees from obligations and removed the clauses from standard paperwork.

Data privacy and GDPR

Italy (December 2024): Garante fined OpenAI €15 million for GDPR violations. Multiple European countries have launched investigations. Austria’s NOYB filed complaints alleging OpenAI cannot correct false information about individuals (hallucinations).

Structural controversy

OpenAI’s December 2024 announcement to convert fully to for-profit sparked backlash from former employees, Elon Musk (who filed suit), and civic leaders. An open letter argued restructuring would “remove crucial governance safeguards” and “abandon founding mission.”

May 2025: OpenAI reversed course after pressure from California and Delaware Attorneys General. The October 2025 restructure kept the nonprofit in control of a Public Benefit Corporation rather than a standard for-profit.

Recent developments (2024-2025)

The corporate restructure (October 2025)

OpenAI completed its restructuring on October 28, 2025:

• The nonprofit became OpenAI Foundation (holds 26% stake worth ~$130B)
• The for-profit arm is now OpenAI Group PBC (Public Benefit Corporation)
• Microsoft holds 27% equity (~$135B value)
• Employees and investors hold 47%
• Microsoft IP rights extended through 2032, including post-AGI models

Stargate and infrastructure

In January 2025, OpenAI announced the Stargate Project at the White House, a $500 billion joint venture with SoftBank and Oracle to build 10GW of AI data centres across the U.S. The first site in Abilene, Texas includes an Oracle order for 400,000 Nvidia GB200 GPUs.

OpenAI is also developing custom AI chips with Broadcom in a $10B deal, targeting 10GW capacity by 2029, with first chips expected Q3 2026 via TSMC fabrication.

”Code red” and competition (December 2025)

Sam Altman reportedly declared “code red” after Google’s Gemini 3 release. OpenAI is delaying monetisation experiments (ads, retail/healthcare agents) to focus on making ChatGPT “more intuitive and personal.” GPT-5.2 is reportedly ready for release around December 9, 2025 to respond to Gemini 3.

Acquisitions and expansion

December 3, 2025: OpenAI announced acquisition of Neptune.ai (experiment tracking). The company launched “OpenAI for Countries” for international deployments and formed SB OpenAI Japan with SoftBank for the Japanese market.

Where OpenAI excels

General-purpose AI assistance

ChatGPT remains the default choice for most users. The combination of brand recognition, free tier access, and GPT-5’s unified architecture makes it the most accessible entry point to AI. 800 million weekly users can’t all be wrong, though they can be habitual.

Multimodal capabilities

OpenAI leads in combining text, image, audio, and video in a single interface. GPT-4o’s native multimodality, DALL-E 3’s image generation, Sora’s video creation, and Whisper’s transcription create an ecosystem no competitor fully matches.

Consumer product polish

The ChatGPT interface, mobile apps, and browser extensions are refined products. Features like conversation memory, custom GPTs, and voice mode demonstrate consumer product thinking that enterprise-focused competitors lack.

Enterprise adoption

92% of Fortune 500 companies use OpenAI products. The Microsoft relationship provides enterprise sales channels, compliance frameworks, and Azure integration that startups cannot match.

Where OpenAI falls short

Coding (versus Claude)

Claude Sonnet 4.5 consistently outperforms GPT models on SWE-bench and real-world coding tasks. Developers increasingly prefer Claude for complex programming, with Anthropic capturing significant developer mindshare despite smaller overall market share.

Context window (versus Gemini)

Gemini offers 1M-2M token context windows versus GPT-4’s 128K. For document analysis, codebase understanding, and long-form content, Google’s advantage is substantial.

Transparency and safety

The exodus of safety researchers, dissolution of the Superalignment team, and weakened safety frameworks have damaged OpenAI’s credibility in the AI safety community. Former employees have been openly critical.

Cost efficiency (versus DeepSeek, Gemini)

DeepSeek offers comparable quality at 10-20% of OpenAI’s API costs. Gemini Flash undercuts GPT-4o mini. For cost-sensitive applications, alternatives increasingly make sense.

Privacy concerns

European regulators have fined OpenAI and launched multiple investigations. The company’s data collection practices remain controversial, particularly regarding training data sources and user conversation handling.

Developer resources

Official documentation

• Platform documentation — API reference, guides, tutorials
• API reference — Endpoint specifications
• Cookbook — Code examples and recipes
• Status page — Service health monitoring

SDKs and libraries

• Python SDK — Official Python client
• Node.js SDK — Official JavaScript/TypeScript client
• .NET SDK — Official C# client

Key tools

• Playground — Interactive model testing
• Fine-tuning — Custom model training
• Assistants API — Stateful agent building
• Batch API — Async bulk processing at 50% discount

FAQ

Is ChatGPT free?

Yes. ChatGPT offers a free tier with access to GPT-4o mini and limited GPT-4o usage. Paid plans ($20-200/month) provide full model access, higher usage limits, and additional features.

How much does the OpenAI API cost?

Pricing varies by model. GPT-4o: $2.50/1M input tokens, $10/1M output. GPT-4o mini: $0.15/$0.60. GPT-4.1: $2/$8. o3: $2/$8. Full pricing at openai.com/api/pricing.

Is OpenAI owned by Microsoft?

No. Microsoft owns 27% equity as of October 2025. The OpenAI Foundation (nonprofit) owns 26%, with employees and other investors holding 47%. Microsoft has significant commercial rights but not control.

What happened with Sam Altman’s firing?

In November 2023, OpenAI’s board fired Altman citing lack of candour. After 97% of employees threatened to quit, Altman returned within five days. The original board members departed, and Altman later rejoined a reconstituted board.

Is OpenAI safe to use for business?

OpenAI offers enterprise tiers with data privacy commitments, SOC 2 compliance, and options for data not used in training. However, European privacy regulators have fined OpenAI, and data handling practices remain under scrutiny.

What’s the difference between GPT-4, GPT-4o, and GPT-5?

GPT-4 (March 2023) was text-focused. GPT-4o (May 2024) added native multimodal capabilities. GPT-5 (August 2025) unified reasoning and fast-response modes with a real-time router. GPT-5 is the current flagship.

Which OpenAI model is best for coding?

For most coding tasks, GPT-4o or GPT-4.1 (1M context) works well. For complex reasoning, o3 excels but is slower and more expensive. Note: Claude Sonnet 4.5 outperforms GPT models on SWE-bench benchmarks.

Can I use ChatGPT for commercial purposes?

Yes. OpenAI’s terms permit commercial use of outputs. API usage requires compliance with usage policies. Enterprise tiers provide additional commercial guarantees and indemnification.