Key Whisper Stats 2025
OpenAI released Whisper in September 2022 as an open-source automatic speech recognition system trained on 680,000 hours of multilingual audio data. The Microsoft-backed company developed this transformer-based encoder-decoder model to convert spoken language into text with near-human accuracy. Today, Whisper powers transcription workflows across 99 languages and processes audio through six model variants ranging from the compact 39-million parameter Tiny version to the full-scale 1.55-billion parameter Large-v3. You can access Whisper through OpenAI’s API at $0.006 per minute or self-host it entirely free under the MIT license.
Parameters in Whisper Large-v3 model
Hours of audio training data for Large-v3
Word error rate for Large-v3-turbo
Monthly downloads on HuggingFace
Cost per minute via OpenAI API
OpenAI Whisper History Timeline
Whisper’s development traces back to OpenAI’s founding mission of building beneficial AI systems. The speech recognition model emerged from years of research into transformer architectures and large-scale audio training. Here’s how Google, Amazon, and other tech giants watched OpenAI disrupt the speech-to-text market with an open-source solution that challenged expensive proprietary alternatives.
Sam Altman, Elon Musk, Greg Brockman, and Ilya Sutskever established OpenAI as a nonprofit research lab with $1 billion in pledged funding from investors including Peter Thiel and Reid Hoffman.
Microsoft invested $1 billion in OpenAI, providing Azure cloud infrastructure that would later power Whisper’s training on massive audio datasets.
OpenAI launched Whisper with six model sizes (Tiny, Base, Small, Medium, Large, Large-v2) trained on 680,000 hours of multilingual audio, supporting 99 languages under the MIT open-source license.
OpenAI released Whisper Large-v3 with 1.55 billion parameters, trained on over 5 million hours of audio data. The update expanded to 100 languages with Cantonese support and improved spectral resolution using 128 Mel frequency bins.
The turbo variant reduced decoder layers from 32 to 4, achieving 8x faster processing while maintaining accuracy comparable to Large-v2. VRAM requirements dropped 40% to approximately 6GB.
Whisper leads HuggingFace’s Open ASR Leaderboard with over 4.29 million monthly downloads. The model powers 1,773+ GitHub repositories across Python, JavaScript, Rust, and Swift implementations.
OpenAI Co-founders Behind Whisper
OpenAI launched in December 2015 with eleven co-founders who believed artificial general intelligence required an open research approach. The founding team combined Silicon Valley entrepreneurs with world-class AI researchers from institutions like Stanford and the University of Toronto. Their vision shaped the company culture that eventually produced Whisper’s breakthrough speech recognition capabilities.
Former Y Combinator president who serves as OpenAI’s CEO. Altman co-chaired the organization alongside Musk during its founding and has led the company’s transformation from research lab to commercial powerhouse.
Tesla and SpaceX CEO who co-chaired OpenAI and contributed $45 million in early funding. Musk departed the board in 2018 due to disagreements over the company’s direction and potential conflicts with Tesla’s AI work.
Former Stripe CTO who became OpenAI’s president and chairman. Brockman recruited the initial research team and played a central role in building the engineering culture that developed Whisper.
Chief scientist who co-invented AlexNet under Geoffrey Hinton at the University of Toronto. Sutskever led OpenAI’s research direction before departing in May 2024 to pursue new ventures.
Reinforcement learning researcher who developed key algorithms used across OpenAI’s products. Schulman left OpenAI in August 2024 to join Anthropic.
Polish mathematician and computer scientist who turned down offers worth two to three times his market value to join OpenAI. He oversees the Codex research and language teams developing GPT successors.
OpenAI Revenue Growth
OpenAI’s revenue trajectory represents one of the fastest growth rates in corporate history. The company grew from $3.5 million in 2020 to over $10 billion in annual recurring revenue by mid-2025. ChatGPT subscriptions drive the majority of income, with 15 million active subscribers paying $20 monthly for ChatGPT Plus access.
The Whisper API contributes to OpenAI’s broader API revenue, which generated approximately $200 million annually through the Microsoft Azure partnership. API access pricing at $0.006 per minute undercuts competitors like Amazon Transcribe, which charges $0.024-$0.036 per minute. This aggressive pricing strategy helped establish Whisper as the default choice for developers building speech-to-text applications.
Whisper Speech Recognition Competitors
The speech-to-text market features established cloud providers and specialized startups competing for enterprise transcription workloads. Google Cloud Speech-to-Text supports 125+ languages with deep integration into Google Workspace. Amazon Transcribe offers real-time processing with seamless AWS ecosystem connectivity. Whisper differentiates through open-source availability and competitive pricing that challenges proprietary alternatives.
Independent benchmarks show AssemblyAI achieving the lowest word error rates on diverse audio samples, while Whisper excels in multilingual accuracy and cost efficiency. The table below compares pricing, language support, and key capabilities across major speech recognition providers.
| Provider | Price/Min | Languages | Key Strength |
|---|---|---|---|
| OpenAI Whisper | $0.006 | 100 | Open-source, lowest API cost |
| Google Cloud STT | $0.006-$0.009 | 125+ | Google Workspace integration |
| Amazon Transcribe | $0.024 | 100+ | AWS ecosystem, call analytics |
| Microsoft Azure Speech | $0.016 | 100+ | Teams/Office 365 integration |
| AssemblyAI | $0.015 | 99 | Highest accuracy, sentiment analysis |
| Deepgram | $0.0043 | 30+ | Low latency, developer-friendly |
| Rev.ai | $0.02 | 36 | Human review option available |
| Speechmatics | Custom | 50 | Accent handling, enterprise focus |
| IBM Watson STT | $0.01 | 17 | Custom vocabulary training |
| NVIDIA Parakeet | Self-host | English | GPU optimization, low deletion errors |
OpenAI Valuation and Market Cap
OpenAI’s valuation skyrocketed from $1 billion in 2019 to $500 billion by October 2025, making it one of the most valuable private companies globally. The company trails only SpaceX ($350 billion) among privately held enterprises. This explosive growth reflects investor confidence in OpenAI’s position as the leading commercial AI developer.
The $40 billion funding round in March 2025, led by SoftBank’s $30 billion commitment, set records for private tech investment. Microsoft‘s cumulative $14 billion investment secured the tech giant a 27% stake following OpenAI’s October 2025 conversion to a for-profit public benefit corporation.
OpenAI Acquisitions
OpenAI’s acquisition strategy shifted dramatically in 2023 when the company completed its first-ever purchase of Global Illumination. Founded by former Instagram and Facebook engineers Thomas Dimson, Taylor Gordon, and Joey Flynn, the New York-based startup specialized in AI-powered creative tools. The entire team joined OpenAI to work on core products including ChatGPT, bringing expertise in consumer technology and product design.
The company accelerated its M&A activity throughout 2024 and 2025. OpenAI acquired Multi, a remote collaboration platform built by former Dropbox and Google engineers, to enhance enterprise collaboration capabilities. Rockset, a real-time analytics database provider, sold to OpenAI for approximately $500 million, strengthening the company’s data infrastructure for training and serving AI models.
The landmark acquisition came in May 2025 when OpenAI purchased io, the AI hardware startup founded by legendary Apple designer Jony Ive, for $6.5 billion. This deal marks OpenAI’s largest acquisition and signals the company’s ambitions beyond software into consumer devices. Ive, whose design credits include the iPhone, iPad, and iMac, now leads creative and design work across all OpenAI operations. The io team of 55 engineers, scientists, and product specialists includes former Apple designers Scott Cannon, Evans Hankey, and Tang Tan.
OpenAI also pursued Windsurf, an AI-assisted coding tool previously known as Codeium, with a reported acquisition price around $3 billion. This purchase enhances OpenAI’s capabilities in providing advanced coding assistance to developers. Additionally, the company secured a $350 million stake in CoreWeave to access AI infrastructure, committing $11.9 billion over five years for computing resources. These strategic moves position OpenAI to compete across software, hardware, and infrastructure segments of the AI market.
FAQs
What is OpenAI Whisper used for?
Whisper converts spoken audio into written text across 100 languages. Developers use it for meeting transcription, podcast processing, voice assistants, call center analytics, and accessibility applications. The model handles translation, timestamps, and language detection in a single system.
Is Whisper free to use?
Yes. OpenAI released Whisper under the MIT open-source license, allowing unlimited commercial and research use. You can self-host the model at no cost or access OpenAI’s API at $0.006 per minute for managed transcription services.
How accurate is Whisper compared to human transcription?
Whisper Large-v3-turbo achieves 7.75% word error rate on clean audio, approaching the 4-6.8% range of professional human transcribers. Performance degrades in noisy environments, where insertion error rates increase significantly compared to competitors.
What hardware do I need to run Whisper locally?
Requirements vary by model size. Tiny runs on 1GB VRAM suitable for mobile devices. Large-v3 needs approximately 10GB VRAM with a powerful GPU. Large-v3-turbo reduces requirements to 6GB while maintaining comparable accuracy.
Who founded OpenAI and when?
Sam Altman, Elon Musk, Greg Brockman, Ilya Sutskever, and seven others founded OpenAI in December 2015. The nonprofit research lab received $1 billion in pledged funding from investors including Peter Thiel, Reid Hoffman, and Amazon Web Services.
