Standard Intelligence
Standard Intelligence is a San Francisco-based AI research company that builds foundation models for general-purpose computer use (FDM-1) and open-source conversational speech (hertz-dev), targeting enterprise automation and developer adoption with proprietary video pretraining infrastructure.
- Company typePrivate
- Founded2024
- HeadquartersSan Francisco, United States
- Headcount11–50
- GTM typeB2B
- OfferingSoftware
What Standard Intelligence does
Standard Intelligence is a San Francisco-based AI research company operating as Standard Intelligence PBC that builds foundation models aimed at general-purpose computer use and interactive audio. Its flagship model, FDM-1, is trained on an 11-million-hour IDM-labeled video corpus and is designed to drive graphical user interfaces directly via key presses, mouse movements, and scroll events, with demonstrated capability in CAD (Blender), web exploration, GUI fuzzing, and, with under an hour of fine-tuning, real-world car driving. The company also open-sources hertz-dev, an 8.5B-parameter autoregressive base model for full-duplex conversational speech that achieves 120ms real-world latency on a single RTX 4090. Underpinning this portfolio is "the heap," a 30-petabyte in-house storage cluster built in downtown San Francisco for under $500K (roughly 40x cheaper than AWS), and an evaluation infrastructure capable of running 1M rollouts per hour across 80,000 forking VMs.
The company's core technical differentiator is a custom self-supervised video encoder that compresses nearly two hours of 30 FPS video into 1M tokens — reported as ~100x more token-efficient than OpenAI's encoder — enabling internet-scale video pretraining where prior computer-use approaches were constrained to tens of hours. The model is trained autoregressively on interleaved video and discrete action tokens, allowing it to function as a "coworker" foundation model for CAD, finance, engineering, and eventually ML research, with a stated long-term goal of superhuman performance on general computer tasks and the development of alignment techniques for general learners.
Standard Intelligence is pre-revenue and pre-commercial. Its current go-to-market is community-led: hertz-dev is distributed as open weights on GitHub and via downloadable checkpoints to build developer mind-share, while FDM-1 is shared only through the company's research blog. No public API, SDK, pricing, or named customers exist as of April 2026. The company anticipates future commercial distribution via enterprise sales for FDM-1, and the $75M Series A closed on April 30, 2026 (co-led by Sequoia Capital and Spark Capital at a ~$500M post-money valuation, with angels Milan Kovac, Stanley Druckenmiller, and Andrej Karpathy) is earmarked for compute scaling, alignment research, and hiring researchers and engineers in San Francisco.
Standard Intelligence firmographics
Firmographics- Name
- Standard Intelligence
- Legal name
- Standard Intelligence PBC
- Website
- https://si.inc
- Company type
- Private
- Founded year
- 2024
- Operating status
- Operating
- Headcount range
- 11–50 employees
- Short description
- Standard Intelligence is a San Francisco-based AI research company that builds foundation models for general-purpose computer use (FDM-1) and open-source conversational speech (hertz-dev), targeting enterprise automation and developer adoption with proprietary video pretraining infrastructure.
- Ownership category
- akta.pro rank
Standard Intelligence industry classification
Industry- Product category
- AI Foundation Models
- NAICS
- Custom Computer Programming Services (541511), Computer Systems Design and Related Services (54151)
- SIC
- Services-Prepackaged Software (7372), Services-Computer Integrated Systems Design (7373), Services-Computer Programming, Data Processing, Etc. (7370)
- akta.pro industry
- Foundation Model Developers (LLM/Multimodal Model Labs)Hardware, Technology & Telecom/Artificial Intelligence & Machine Learning/Foundation Models & Generative AI (LLMs, multimodal)
Keywords
Where Standard Intelligence is headquartered
LocationHeadquarters
- HQ city
- San Francisco
- HQ country
- United States
- HQ region
- North America
Offices1 record
Markets served
Standard Intelligence business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Personnel, Technology or R&D, Infrastructure, Marketing or Sales
Revenue model
- Pre-revenue / Research stage: Company is pre-revenue and pre-commercial; no disclosed revenue-generating products or services. Funding is from a $75M Series A announced in May 2026 at approximately $500M post-money valuation. Future revenue model is not publicly disclosed but likely involves API access, licensing, or enterprise deployment of the FDM-1 computer action foundation model.
Go-to-market motion2 records
Distribution channels2 records
Marketing channels3 records
Standard Intelligence product offering
Product offeringCore offering
Standard Intelligence develops AI foundation models, anchored by FDM-1, a foundation model for computer use trained on an 11-million-hour screen-recording video corpus that predicts key presses and mouse actions to automate any GUI-based workflow, and by hertz-dev, an 8.5B-parameter open-source autoregressive base model for full-duplex conversational speech. The company is currently pre-revenue and distributes hertz-dev via open-source channels while positioning FDM-1 for future enterprise deployment, supported by a proprietary 30-petabyte in-house storage cluster and an evaluation infrastructure capable of 1 million rollouts per hour.
Product overview
Standard Intelligence is an AI research company that offers a portfolio of foundation-model products rather than a single unified SaaS application. Its flagship research offering is FDM-1, a foundation model for computer use that learns from an 11-million-hour video corpus to automate any GUI-based workflow; its open-source offering is hertz-dev, an 8.5B-parameter base model for full-duplex conversational speech, composed of the hertz-codec audio VAE and the hertz-ar autoregressive transformer. Underpinning this portfolio is "the heap," a 30-petabyte in-house storage cluster built to make internet-scale video pretraining economically viable. The company's stated long-term goal is to build aligned general learners capable of superhuman computer use, with the two product lines (FDM-1 and hertz-dev) representing different pillars of that research agenda.
Differentiator
Problem solved
Functional benefit
Products and services
- FDM-1 (Forward Dynamics Model 1) FDM-1 is a foundation model for computer use that ingests screen-recording video at 30 FPS and autoregressively predicts the next computer action (key presses, mouse movements, scroll events). It is designed to automate any GUI-based workflow and serves as a coworker model for CAD, finance, engineering, and eventually ML research, with demonstrated capabilities in CAD modeling in Blender, GUI fuzzing of banking apps, and autonomous driving via key presses.
- hertz-dev hertz-dev is an 8.5B-parameter open-source autoregressive base model for full-duplex, two-speaker conversational speech generation. It natively supports overlapping audio, interruptions, and faster-than-human reaction times, targeting developers and researchers building interactive voice and conversational AI applications.
Quantifiable outcome
- Video encoder 100x more efficient than OpenAI's
- +4 more outcomes
Companies that use Standard Intelligence
Customer profileSegments2 records
Ideal customer profiles2 records
Standard Intelligence technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- No
- API docs
- API detail
Core technology
AI maturity
App detail
AI capability13 records
Feature5 records
Standard Intelligence partnerships and signals
Strategic signalScale indicators12 records
Recent moves6 records
Expansion highlights5 records
Standard Intelligence competitors and assessment
Company assessmentBroad incumbents
- OpenAI: OpenAI ships general foundation models (GPT, Operator/Computer Use, video generation) and is the primary benchmark for Standard Intelligence's '100x more efficient than OpenAI' video encoder claim. Operates across language, vision, audio, and computer-use, making it the closest broad incumbent peer.
- Google DeepMind: DeepMind develops frontier multimodal and agentic foundation models (Gemini, robotics, video) and has historically led on video pretraining research, making it a key comparator for FDM-1's video-pretraining recipe and long-context computer-use claims.
- Anthropic: Anthropic is a frontier foundation-model lab whose Claude 'Computer Use' capability directly competes with FDM-1 in the GUI-automation category, and it ships conversational AI and enterprise APIs comparable to the future commercialization path of Standard Intelligence's models.
- Mistral AI: Mistral AI is a well-funded European foundation-model lab shipping open-weight language and multimodal models, comparable to the open-source hertz-dev release and to the future commercial/API offering Standard Intelligence intends to build on top of FDM-1.
Direct peers
- Adept AI: Adept is building foundation-model 'AI agents' that operate software via graphical interfaces, the same product category as FDM-1, making it the most directly comparable startup peer. Both target enterprise software automation through visual interaction with GUIs.
- ElevenLabs: ElevenLabs is a leading commercial conversational-speech and voice-cloning AI company, directly comparable to the hertz-dev open base model for low-latency full-duplex speech generation. Both target low-latency, high-quality voice AI applications.
- Sierra: Sierra builds conversational AI agents for enterprise customer experience, comparable to the enterprise-automation use cases Standard Intelligence targets with FDM-1. Both position foundation-model agents as 'coworkers' for enterprise workflows.
Others
- Hugging Face: Hugging Face is the dominant open-source model hub and distribution platform, which is the channel through which hertz-dev-style open models reach developers. It is adjacent rather than directly competitive, but a key ecosystem participant for Standard Intelligence's open-source GTM motion.
Emerging players
- MultiOn: MultiOn is an early-stage agentic-AI startup building autonomous web-browsing and task-completion agents, overlapping with FDM-1's GUI-automation use cases though with a narrower focus on consumer/prosumer web tasks.
- Reka AI: Reka builds multimodal foundation models (text, image, audio, video) for enterprise deployment, comparable to Standard Intelligence's multimodal FDM-1/hertz-dev stack. Both are privately held multimodal model labs targeting enterprise applications.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks7 records
Key highlights7 records
Customer concentration
Standard Intelligence social profiles
Digital presenceStandard Intelligence financial estimates
Financial estimateRevenue estimate
Valuation estimate
Standard Intelligence leadership team
Management profileNumber of profiles
Profiles4 records
Standard Intelligence funding detail
Funding detailFunding overview
Funding rounds1 record
Investors2 records
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Standard Intelligence M&A and investment
M&A and investmentM&A
Investments
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Standard Intelligence
What does Standard Intelligence do?
Standard Intelligence develops AI foundation models, anchored by FDM-1, a foundation model for computer use trained on an 11-million-hour screen-recording video corpus that predicts key presses and mouse actions to automate any GUI-based workflow, and by hertz-dev, an 8.5B-parameter open-source autoregressive base model for full-duplex conversational speech. The company is currently pre-revenue and distributes hertz-dev via open-source channels while positioning FDM-1 for future enterprise deployment, supported by a proprietary 30-petabyte in-house storage cluster and an evaluation infrastructure capable of 1 million rollouts per hour.
Is Standard Intelligence a public or private company?
Standard Intelligence is a private company. It is classified as venture growth investor backed and is currently operating.
When was Standard Intelligence founded?
Standard Intelligence was founded in 2024. It employs 11 to 50 people.
Where is Standard Intelligence based?
Standard Intelligence is headquartered in San Francisco, United States, in the North America region.
How does Standard Intelligence make money?
One revenue line is on record: pre-revenue / Research stage.
Who are Standard Intelligence's main competitors?
Broad incumbents on record are OpenAI, Google DeepMind, Anthropic and Mistral AI. Direct peers are Adept AI, ElevenLabs and Sierra. Hugging Face is listed as an others. Emerging players are MultiOn and Reka AI.
Does Standard Intelligence have an API?
No public API is recorded for Standard Intelligence.
What industry is Standard Intelligence in?
Standard Intelligence's product category is AI Foundation Models. Its primary akta.pro industry code is HDAAACAA, Foundation Model Developers (LLM/Multimodal Model Labs). Its NAICS code is 541511 and its SIC code is 7372.