What Sapient Intelligence does

Sapient Intelligence is a privately held AI research company founded in August 2024 and incorporated in Singapore as Sapient Intelligence Pte. Ltd., with operating presence in Singapore, Beijing, and Palo Alto. The company is pursuing Artificial General Intelligence (AGI) by developing a proprietary brain-inspired architecture called the Hierarchical Reasoning Model (HRM), which performs reasoning in continuous latent space rather than through token-level chain-of-thought. HRM is structured as a two-level recurrent system: a high-level "Slower Controller" handles abstract, deliberate reasoning while a low-level "Faster Processor" performs detailed computation, with cross-level information flow at every recurrent step. The architecture has been published as a research paper (June 2025) and productized into two open-source models — HRM-Symbolic (27 million parameters, released July 2025) for structured puzzle and ARC-AGI reasoning, and HRM-Text (1.15 billion parameters, released May 2026) for language tasks — both distributed freely on GitHub and Hugging Face.

The company's product portfolio consists of the underlying HRM architecture and the two named models built on top of it, with an explicit roadmap toward model-as-a-service APIs, fine-tuning consoles, agentic AI Scientist tooling, and pricing tiers. There are no disclosed commercial customers, no published API endpoints or pricing, and the Terms & Conditions state that current interactive demonstrations are not commercial products. Revenue today is effectively zero on the model side, with open-source distribution functioning as a developer-adoption and brand-building mechanism rather than a monetization channel. The latent option to license HRM technology is reserved to the Singapore parent, providing a future path for proprietary or dual-licensing revenue.

Go-to-market is community-led via GitHub, Hugging Face, Discord, X, YouTube, and active hiring (five public roles), targeting verticals including Embodied AI, Healthcare, Quantitative Finance, Climate, and AI4S. The company is venture-backed, having closed a $22 million seed round in late 2024 with Minerva Capital and reached a post-money valuation above $200 million after three funding rounds within six months of founding; founder Guan Wang (CEO) and William Chen (Head of Strategy & Management), both Tsinghua University graduates, retain operational control. Customers today are open-source developers and researchers rather than paying enterprises, and the business model is pre-revenue with commercialization planned via future APIs.

Name

Sapient Intelligence

Legal name

Sapient Intelligence Pte. Ltd.

Website

https://sapient.inc

Company type

Private

Founded year

2024

Operating status

Operating

Headcount range

11–50 employees

Short description

Sapient Intelligence is a Singapore-incorporated AI research lab developing the proprietary Hierarchical Reasoning Model (HRM), a brain-inspired latent-space architecture open-sourced as HRM-Symbolic and HRM-Text to build developer adoption ahead of planned model-as-a-service commercialization.

Ownership category

akta.pro rank

Product category

AI Foundation Models

NAICS

Custom Computer Programming Services (541511), Software Publishers (5132), 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 primary industry

Foundation Model Developers (LLM/Multimodal Model Labs) (HDAAACAA)

akta.pro secondary industries

Model Development & Training Platforms (AutoML, Notebooks, Feature Stores) (HDAEANAB), Enterprise AI Data & Knowledge Platforms (Vector Databases, Knowledge Graphs) (HDAEANAH)

Keywords

• Hierarchical reasoning models
• Brain-inspired AI architecture
• Latent space reasoning
• Open source AI models
• Foundation model development

Headquarters

HQ city

Beijing

HQ country

China

HQ region

Asia

Markets served

GTM type

B2B

Offering type

Software

Cost components

Personnel, Technology or R&D, Infrastructure, Marketing or Sales, Operations

Revenue model

1. Open-source release (current): Currently pre-revenue on the model side. HRM-Symbolic (open-sourced July 2025) and HRM-Text (open-sourced May 2026) are released under open-source licenses on GitHub and Hugging Face with weights and code freely available, generating no direct revenue but building developer adoption, brand, and ecosystem.
2. Planned model-as-a-service APIs and fine-tuning platform (future): Forward-looking commercialization: PM hiring scope explicitly covers model-as-a-service APIs, fine-tuning consoles, agentic workflow tools, and pricing tiers / hardware-aware monetization. Indicates intent to layer API/usage-based revenue on top of the open-source base.
3. Licensing of HRM technology (latent): Terms reserve all rights in the HRM technology, architectural diagrams, and brand to Sapient Intelligence Pte. Ltd. Open-source code on GitHub is governed by separate open-source licenses. Provides an option for future proprietary/dual licensing of architecture, models, or branding.

Pricing tiers

Go-to-market motion1 record

Core offering

Sapient Intelligence develops a brain-inspired AI architecture called the Hierarchical Reasoning Model (HRM) and ships it as open-source models for complex reasoning tasks. The current product portfolio consists of HRM-Symbolic (a 27M-parameter reasoning model optimized for structured problem-solving such as Sudoku, ARC-AGI, and mazes) and HRM-Text (a 1.15B-parameter text generation model). Both are released under open-source licenses on GitHub and Hugging Face.

Product overview

Sapient Intelligence offers a unified brain-inspired AI architecture brand — the Hierarchical Reasoning Model (HRM) — packaged as two distinct open-source model products that share the same underlying HRM latent-space hierarchical recurrent architecture. HRM-Symbolic is the original reasoning model optimized for compact, structured problem-solving on Sudoku, ARC-AGI, and Maze benchmarks with only 27M parameters and edge-deployable inference. HRM-Text is the text-generation extension of the same architecture at 1.15B parameters, trained on ~40B tokens for ~$1,000, with a 0.6 GiB int4 footprint enabling on-device reasoning. Together these two named models (plus the underlying HRM architecture itself) form the company's current product portfolio, with the company pursuing Artificial General Intelligence (AGI) and building toward model-as-a-service APIs, fine-tuning platforms, and agentic tooling for AI Scientists.

Differentiator

Problem solved

Functional benefit

Brands

• HRM-Symbolic: A reasoning model designed for deep, efficient problem-solving through hierarchical modules, dual timescale computation, and latent space reasoning; open-sourced in July 2025.
• HRM-Text

Products and services

• HRM (Hierarchical Reasoning Model) Architecture The core architecture underlying Sapient Intelligence's product portfolio: a brain-inspired latent-space hierarchical recurrent model with two-timescale computation that performs reasoning in continuous latent space rather than via discrete language tokens. Sold as an open architecture referenced from the company's product pages and GitHub repositories for use by AI researchers and developers building reasoning systems.
• HRM-Symbolic A 27-million-parameter reasoning model built on the HRM architecture for deep, efficient problem-solving on structured tasks such as Sudoku, ARC-AGI, and mazes. Achieves state-of-the-art benchmark performance without pretraining or chain-of-thought supervision, supports small-sample learning, and is 100x faster than comparable LLM approaches. Distributed free under open-source license for AI researchers and developers.
• HRM-Text A 1.15-billion-parameter text generation model based on the HRM architecture, strengthened by task-completion training and latent-space reasoning. Achieves 56.2% MATH, 82.2% DROP, 81.9% ARC-Challenge, and 60.7% MMLU benchmarks while supporting native edge reasoning with a 0.6 GiB int4 footprint. Distributed free under open-source license for on-device deployment.

Ideal customer profiles2 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration2 records

AI capability6 records

Feature7 records

Recent moves6 records

Expansion highlights5 records

Others

• xAI:
• Hugging Face: AI platform that hosts Sapient's open-source model weights and is a critical distribution and developer-traction channel for the company. Not a competitor but a key ecosystem enabler comparable in role to a platform partner.

Direct peers

• DeepSeek: Chinese AI lab that has open-sourced highly efficient reasoning-focused foundation models (DeepSeek-R1, V3) at substantially lower training cost than US frontier labs. Directly comparable to Sapient on geography, open-source distribution, and cost-efficient reasoning focus.
• Mistral AI: European AI lab that built a multi-billion-dollar valuation by open-sourcing efficient small-to-mid-size foundation models and layering commercial API and fine-tuning on top. Closest analog to Sapient's open-source-first + planned-API-monetization GTM.
• Moonshot AI: Chinese foundation model lab (Kimi) pursuing long-context and reasoning capabilities with significant capital. Comparable to Sapient in Chinese origin, foundation-model focus, and reasoning-product positioning, though at much larger scale.
• Sakana AI: Tokyo-based AI research lab founded by ex-Google Brain researchers that pursues nature-inspired and evolutionary approaches to foundation models. Direct architectural parallel to Sapient's brain-inspired, non-Transformer reasoning research.
• Liquid AI: MIT-spinout building efficient, non-Transformer foundation models (Liquid Foundation Models) with state-space and other novel architectures. Comparable as a research-stage, well-funded non-Transformer architecture company chasing reasoning and efficiency.

Broad incumbents

• Anthropic: Frontier AI lab with leading reasoning models (Claude 3.7 Sonnet, Claude 4) and a focus on chain-of-thought reasoning and safety. The incumbent Sapient positions HRM against on long-horizon reasoning tasks.
• OpenAI: Foundation model leader with the o-series reasoning models (o1, o3, o4-mini) that define the chain-of-thought reasoning category Sapient's latent-space approach is explicitly positioned as an alternative to.

Emerging players

• Cartesia: Early-stage AI lab building state-space models (Mamba) as an efficient non-Transformer alternative for sequence modeling. Architecturally adjacent to Sapient's pursuit of alternatives to Transformer-based LLMs.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat5 records

Key highlights7 records

Customer concentration

• Xx.com/Sapient_Int
• GitHubgithub.com/sapientinc
• YouTubeyoutube.com/channel/UCd1pXtPtluwfKCRZCgKL63w
• LinkedInlinkedin.com/company/sapientinc

Revenue estimate

Valuation estimate

Number of profiles

Profiles2 records

Funding overview

Funding rounds2 records

Investors2 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Sapient Intelligence