Sentient is a San Francisco-based technology company focused on developing open-source Artificial General Intelligence (AGI). ^[6] The organization's primary project is "the GRID," an open intelligence network designed to orchestrate a multitude of AI models, agents, data, and tools. The ecosystem is coordinated through its native token, SENT. ^{[1] [3] [3]}​

Overview

Sentient's stated mission is to ensure that Artificial General Intelligence remains open-source and is not controlled by any single entity. ^[7] The organization's vision is to close the "core reasoning gap" found in contemporary AI systems, which struggle with tasks requiring sustained, multi-step reasoning under ambiguous conditions. To achieve this, Sentient develops platforms like the Sentient Arena, which creates a structured environment for builders to explore diverse solutions to complex AI problems simultaneously. ^[4]

The organization operates on the principle that an open development model will ultimately surpass the progress of closed, proprietary AI labs. In line with this philosophy, Sentient releases all of its research and technological breakthroughs as open-source contributions. A central tenet of its approach is expressed in its official communications: "Every breakthrough is released open-source, driving our mission and proving openness outpaces closed labs at every step." ^[1]

The core of Sentient's work is the development of "the GRID," which it describes as an open intelligence network. This network is designed to serve as a foundational infrastructure for orchestrating a wide array of AI components, including over 100 models, agents, data sources, and tools, enabling them to function as a cohesive system. The project, based in San Francisco, California, was in a BETA phase as of September 2025. ^{[7] [1]}

History

The organization formally introduced its mission and projects with a blog post titled "What is Sentient" on June 3, 2025.

Throughout mid-2025, Sentient published a series of research papers and articles detailing its work on AI coding capabilities, agent security vulnerabilities, and the architecture of its core project, the GRID. Key technology announcements during this period included the introduction of the ROMA framework on September 8, 2025. ^{[1] [3]}​

In September 2025, the project reported a period of significant activity and growth. On September 16, Sentient announced a major upgrade to the GRID and a new feature for its Sentient Chat platform called "Spaces." The following day, the company reported onboarding 15,000 new users to Sentient Chat within a 12-hour period. This was followed by the announcement of a partnership with Tria for agent payments on September 18 and its planned participation in Korea Blockchain Week later that month. ^[3]​

Technology

Sentient's development efforts cover a full stack of AI technologies, from foundational models and agentic frameworks to security protocols. ^[1]

The GRID

The GRID, which stands for Global Research and Intelligence Directory, is Sentient's central project. It is described as a distributed network of specialized agents, models, data, tools, and compute designed to function as a "distributed brain." Its goal is to coordinate these modular and composable components, known as "Artifacts," to achieve AGI-level intelligence. The system's collective intelligence is designed to compound as more builders contribute to the ecosystem. On November 12, 2025, Sentient announced a further upgrade to the GRID. ^{[8] [7] [12]}

Users access the GRID through interfaces like Sentient Chat, where a query can trigger a coordinated workflow chaining multiple agents and tools to generate an answer. The network integrates contributions from over 100 partners. Builders are incentivized to contribute Artifacts to the GRID, as they can monetize their creations by serving user requests and earning rewards in SENT tokens. The community can stake SENT on valuable Artifacts to help direct rewards and earn yield based on metrics like usage and revenue. ^[7]

Sentient Chat

Sentient Chat serves as the primary public interface and gateway to the GRID. The platform allows users to interact with technologies developed within the ecosystem, test open-source repositories, and provide feedback. It is described as the infrastructure that connects the various elements of the GRID. ^{[1] [3]}

A key feature introduced in September 2025 is "Spaces," designed to streamline user workflows by consolidating information gathering and tools into a single interface, reducing the need for multiple browser tabs. Spaces also incorporates AI-powered analysis, providing users with additional context when they interact with different modules within the interface. ^[3]

ROMA (Recursive Open Meta Agent)

ROMA (Recursive Open Meta-Agent) is an open-source framework released on November 12, 2025, for building high-performance, multi-agent systems. It is designed to overcome the compounding errors found in single-step agents by orchestrating simpler agents and tools to solve complex, long-horizon tasks. The framework is available on GitHub for builders and researchers. ^[13]

ROMA's architecture utilizes a hierarchical, recursive task tree. A parent node breaks a complex goal into smaller subtasks, which are then passed to child nodes for execution. The results from these child nodes flow back up the tree, where the parent node aggregates them into a final solution. The architectural flow is managed by four types of nodes: the Atomizer assesses task complexity, the Planner decomposes complex tasks into subtasks, the Executor carries out simple subtasks, and the Aggregator synthesizes the final answer from child node outputs. ^[13]

Key features of the ROMA framework include transparency through structured inputs and outputs ("stage tracing"), modularity that allows for plugging in any agent, tool, or human-in-the-loop checkpoint, and natural support for parallel execution of tasks. This design also facilitates human-in-the-loop verification at any node, which is useful for very long tasks. ^[13]

To demonstrate its efficacy, Sentient built ROMA Search, an internet search agent. On the SEALQA benchmark, which tests complex, multi-source reasoning, ROMA Search achieved 45.6% accuracy, establishing a new state-of-the-art performance. This surpassed previous systems, including Kimi Researcher (36%), Gemini 2.5 Pro (19.8%), and the next-best open-source system, Open Deep Search (8.9%). ^[13]

SERA (Semantic Embedding & Reasoning Agent)

SERA is an agent architecture designed for efficiency and consistency by using semantic embeddings for routing and reasoning, rather than relying solely on a Large Language Model (LLM). In this framework, an input query is embedded and compared against separate indexes for tools and prompt templates. This allows SERA to select the necessary tools and reasoning prompts in parallel, achieving low latency. ^[5]

An implementation of this architecture, SERA-Crypto, is built on an open-source model stack using Qwen3-Embedding-8B. It is specialized for cryptocurrency analysis and uses a tool index with over 50 endpoints (market data, on-chain flows, etc.) and a prompt index for 11 crypto-specific query types. On Sentient's internal crypto analysis benchmark, SERA-Crypto reportedly ranked #1, outperforming systems such as GPT-5 Medium Reasoning, Grok 4, and Perplexity Finance. ^[5]

Dobby

Dobby is an AI model described by Sentient as the "world’s first fingerprinted AI model." It was developed with a community co-ownership model, reportedly involving over 700,000 individuals. The model was explicitly trained and aligned with values described as "pro-crypto and pro-freedom." Its development and principles are detailed in a publication titled "The Dobby Report." ^[1]

SENT Tokenomics

The Sentient network is coordinated by its native token, SENT, which serves as the coordination layer for the chain, the GRID, and its incentive systems. The total supply of SENT is 34,359,738,368. ^[3]

Allocation

The SENT token distribution is designed with a community-first approach, with the majority of the supply allocated to users and builders. ^[3]

Category	Percentage
Community Allocation	65.55%
↳ Community Initiatives/Airdrop	44.00%
↳ Ecosystem and R&D	19.55%
↳ Public Sale	2.00%
Team	22.00%
Investors	12.45%

Vesting Schedule

To align with the project's long-term goals, founder and investor tokens are subject to extended vesting schedules. ^[3]

• Team: 1-year cliff, followed by a 6-year linear vesting period.
• Investors: 1-year cliff, followed by a 4-year linear vesting period.
• Community Initiatives/Airdrop: 30% unlocked at the Token Generation Event (TGE), with the remainder vesting linearly over 4 years. The airdrop portion is fully unlocked at TGE.
• Ecosystem and R&D: 30% unlocked at TGE, with the remainder vesting linearly over 4 years.
• Public Sale: 100% unlocked at TGE.

Token Utility

The SENT token has compounding utility within the ecosystem. ^[3]

• Staking and Governance: Staked SENT provides voting power in the Sentient DAO, allowing community members to influence decisions on treasury spending, protocol upgrades, and emissions. Staking also provides access to network artifacts like models and data.
• Fees and Payments: SENT functions as the primary currency for services on the network. It is used for payments between agents, models, and data services, enabling artifacts to transact with each other.

Emissions

Token emissions are capped at 2% of the total supply annually to ensure predictability and preserve long-term resources. These emissions are directed to a Community Emission Pool to fund rewards and protocol incentives. Any unspent funds in the pool are locked at the end of each year. ^[3]

Airdrop

In late 2025, Sentient announced an airdrop to reward early contributors and community members. The registration period for the airdrop closed on November 29, 2025. To prevent Sybil attacks, the registration process required a "Billions proof of uniqueness" check, a one-time selfie used to create a zero-knowledge proof without storing the image. The announcement also noted that a second season of contributions was active for future rewards. ^[6]

Eligibility Criteria

Eligibility for the airdrop was extended to several groups based on their contributions to the ecosystem. ^[6]

• Sentient Community: Approximately 1,067 members who earned qualifying roles (Early, Advanced, Sentient AGI) during the first season of community participation. These roles were awarded for evaluated contributions in areas like development, education, and community support.
• Sentient Chat Users: Users were scored based on providing genuine and helpful queries, consistent activity, and valuable feedback or bug reports. A three-layer validation process involving an internal formula, AI grading, and human sampling was used to filter out low-quality interactions.
• Top Voices of Sentient: This group included individuals recognized for their long-term, genuine contributions and influence across social media, evaluated by metrics such as mindshare, tweet volume, and engagement. A portion of the allocation was also designated for top stakers on the Kaito platform.
• Open-Source Researchers: Representing approximately 2% of the total airdrop, this category included external researchers who had collaborated with Sentient, reviewed drafts, contributed datasets, or built upon the project's open-source work.

Research and Publications

Sentient actively publishes its research findings to support its open-source mission. The organization's work focuses on enhancing model capabilities, security, and performance evaluation. ^[1]

Open Deep Search

Sentient developed Open Deep Search, a system that combines open-source Large Language Models (LLMs) with reasoning agents and a proprietary search tool. The organization published research claiming this system achieves performance superior to the GPT-4o Search Preview. The associated paper is available on arXiv. ^{[1] [7]}

Fingerprinting

To address model security and provenance, Sentient created a cryptographic library for "fingerprinting" AI models. This technology embeds undetectable and verifiable digital signatures directly into a model's parameters, which allows for the secure tracking of a model's origin and ownership. ^{[1] [7]}

Benchmarking

The organization develops custom, high-fidelity benchmarks to evaluate AI performance in specialized domains that may be overlooked by general-purpose evaluations. These benchmarks focus on areas such as fair coding practices, alignment with crypto-centric values, and governance principles. Sentient has also developed an internal live crypto analysis benchmark for evaluating token-research agents. ^{[1] [5]}

Agent Security

Sentient's research into agent security has identified significant vulnerabilities within existing open-source agent frameworks. A key area of focus has been "Context Manipulation Attacks," where external inputs can be manipulated to compromise an agent's behavior. The research also demonstrated that specifically fine-tuned models can be more resilient to such attacks and can enhance overall agent security. ^[1]

Partnerships

Sentient has established partnerships to expand the capabilities and ecosystem of the GRID.

In September 2025, Sentient announced a partnership with Tria to serve as the agent payment partner for the GRID. Tria's infrastructure, including its "BestPath" technology, is intended to facilitate seamless financial settlements for AI agents, humans, and other entities on the network. This integration allows agents to trade, transfer, or spend assets across various blockchain and traditional financial systems. ^[3]

The organization is also associated with ecosystem initiatives such as @SentientEco and the @OpenAGISummit, indicating broader community and event involvement. ^[3]

Sentient Foundation

The Sentient Foundation, a global non-profit organization, launched on February 19, 2026. Its mission is to ensure that Artificial General Intelligence remains open-source, decentralized, and aligned with human interests, preventing its control by any single entity. The Foundation acts as a neutral steward, operating alongside Sentient Labs, which functions as the technical research and product organization. ^[9]​

The Foundation's mandate includes establishing safety and alignment standards for AGI, fostering collaboration with researchers, academic institutions, and governments, distributing grants to the open-source developer community, coordinating inclusive governance frameworks, and leading public advocacy for open-source AGI. ^[9]​

Expansion into Asia

The Foundation is actively working to expand its presence in Asia to advance open and collaborative AGI development. The strategy involves building partnerships with universities, researchers, and local developer communities, with a particular focus on East Asia. This initiative aims to democratize AGI innovation by engaging students and regional talent. ^[10]​

In October 2025, Sentient conducted the Sentient East Asia Tour, an outreach program hosted with TinTinLand. The tour included Open AGI Symposiums at Tsinghua Science Park in Beijing on October 24, Shanghai Jiaotong University on October 26, and in Seoul, South Korea, on October 28. Prior campus tours at Fudan University and Shenzhen University drew over 1,500 participants. ^[10]​

Be the Sparks Initiative

On January 2, 2026, the Foundation introduced the Sentient Sparks initiative, a community-driven contributor program designed to empower individuals to take a strategic role in growing the Sentient ecosystem. The program selects ten "Sentient Sparks" each month, composed of six global contributors and four regional contributors from key markets. ^[11]​

Participants are tasked with creating original content, educating newcomers about open-source AI, building online and in-person communities, covering project updates, and co-hosting events. The program's goal is to expand the project's global footprint and accelerate the deployment of open-source AI through a dedicated community of advocates. ^[11]​

Team and Investors

According to its official website, Sentient's research team is composed of individuals from leading universities and AI laboratories, though specific members are not publicly named. ^[1]

The project has received backing from several prominent individual and institutional investors in the technology and cryptocurrency sectors. Notable individual investors include Naval Ravikant and Balaji Srinivasan. Institutional investors include Framework Ventures, Polychain Capital, 1kx, Robot Ventures, Bain Capital Crypto, Nascent, Figment Capital, dao5, Anagram, Archetype, Collab+Currency, Symbolic Capital, Founders Fund, and Pantera Capital. ^{[1] [10]}