# Sentient

> Open-source protocol for community-owned AGI with cryptographic model ownership (OML), multi-agent framework (ROMA), and The GRID network. Backed by an $85M VC seed, still pre-revenue.

- Scores: Freedom 4.7/10 (grade D), Returns 4.8/10 (grade D)
- Category: inference · Token: SENT
- Published: 2026-03-06 · Updated: 2026-10-07
- Canonical: https://ownyourmind.ai/projects/sentient/
- Data (scores + integrity verdicts, JSON): https://ownyourmind.ai/api/projects/sentient.json
- Cite as: Own Your Mind (ownyourmind.ai), licence CC BY 4.0. Link the canonical URL.

Charts and interactive widgets are omitted here; their data is in the JSON above.

---

## What it does

Sentient is building what it calls "community-owned AGI": an open-source protocol where AI model creators can distribute models openly while retaining ownership, earning revenue per use, and enforcing usage rules through cryptography rather than legal contracts.

The core innovation is **OML (Open, Monetizable, Loyal)**, a cryptographic standard published as a peer-reviewed academic paper. OML embeds undetectable fingerprints into language models via fine-tuning. These fingerprints are special (query, response) pairs that uniquely identify model ownership. If someone copies your model, you can prove it cryptographically. If someone uses your model, you earn per-invocation fees via smart contracts.

Three main products:

- **OML Fingerprinting** – The cryptographic model ownership layer. Embed undetectable identifiers into LLMs. Verify ownership without centralised registries. Published at ePrint 2024/1573.
- **ROMA (Recursive-Open-Meta-Agent)** – A multi-agent framework using recursive hierarchical structures for complex task decomposition. Agents call other agents in structured hierarchies. Over 5,000 GitHub stars.
- **The GRID** – An open-source AGI network launched in testnet August 2025. At launch: 40+ AI agents, 50+ data sources, 10+ models deployed across Base, Polygon, and Arbitrum. Users access via Sentient Chat. GRID mainnet has no committed launch date; the network remains in a curated phase.

Founded by four co-founders with an unusual mix of credentials. **Pramod Viswanath** is the Forrest G. Hamrick Professor of Engineering at Princeton and co-inventor of Flash OFDM, the technology underlying 4G mobile networks. **Himanshu Tyagi** is a Professor at the Indian Institute of Science with expertise in privacy and decentralised algorithms. **Sandeep Nailwal** co-founded Polygon (MATIC). **Kenzi Wang** founded Sensys and co-founded Symbolic Capital.

The Sentient Foundation is a nonprofit stewarding the protocol. Sentient Labs handles development. The token generation event concluded on 22 November 2025; the Binance Wallet public sale ran in January 2026 and spot trading opened on 22 January 2026.

## The product in practice

Grounded in our Sentient review, its GitHub and HuggingFace repos, and community signal. Sentient is mostly vision today, so the honest read is about what is and isn't live.

**Who it's for.** Right now, two small groups: people trying the Dobby models (Llama fine-tunes with a crypto-freedom persona, downloadable on HuggingFace) or the hosted Sentient Chat, and developers experimenting with its open agent frameworks like ROMA. The pitch aimed at model creators, own your model and earn per use through OML fingerprinting, has no live marketplace behind it yet. So the addressable user today is a curious tester rather than a builder with a product to ship on it.

**What's actually live.** The shipped pieces are a hosted chat app, a set of open-weight fine-tunes, and open-source agent code. The economic layer that would make Sentient a network rather than a model lab, the GRID with staking and per-invocation fees, launched only in testnet and has no committed mainnet date, so the "community-owned AGI" claim is still a design rather than a running system. Inference for the hosted products runs on a centralised provider rather than a decentralised network. The token trades, but its headline utility depends on that unshipped GRID economy.

**Demand, honestly.** What traction there was came from a burst of hosted engagement around the early-2025 chat launches rather than sustained product use. Model downloads are minuscule for a project this well-funded, and the flagship ROMA agent repo has had no commits since February 2026, which for a developer tool is the most telling signal of all. There is no protocol revenue yet, by the team's own account.

**The vision.** Sentient's idea is ambitious: cryptographically-owned open models via OML, loyalty and provenance baked in, coordinated through a community-owned agent network. The gap is execution, since the ownership and monetisation mechanisms are papers and testnets while the shipped product is a hosted chatbot on someone else's GPUs. Judge it when GRID mainnet and working per-use payments exist; until then the thesis is a bet on the roadmap.

## Value proposition

Three things make Sentient distinctive in crypto-AI.

**Academic rigour.** OML is a peer-reviewed cryptographic protocol with published academic backing. Fingerprinting AI models so that ownership is provable without centralised registries is a genuine contribution. Most crypto-AI projects describe capabilities. Sentient published a paper that proves theirs.

**VC conviction at scale.** $85 million in seed funding led by Founders Fund (Peter Thiel), Pantera Capital, and Framework Ventures, with Franklin Templeton named by Sentient as an ecosystem participant rather than an investor, in a later strategic round. One of the largest seed rounds in crypto history. Twenty-three institutional investors at a $1.2 billion seed valuation.

**65.55% community-facing token allocation.** That's 44% Community + 19.55% Ecosystem + 2% Public. The team's 22% vests over six years, the longest team vesting period in DeAI. The structure puts community ownership in the cap table, not just the marketing.

The counter-narrative is straightforward: **none of this is live yet.** GRID mainnet has no committed launch date. Per-invocation fees are theoretical. There's zero on-chain revenue. At a significant decline from its listing price, SENT trades at an FDV below the seed valuation ($1.2 billion), which means the public market values Sentient lower than what VCs paid, and those VCs' cliff unlock begins in November 2026.

Sentient reports 2 million+ waitlist signups and 100,000+ queries during testing. That suggests meaningful demand, at least at the sign-up stage. But converting a waitlist into paying users for decentralised AI inference is the unsolved problem.

For the sovereignty thesis, OML is architecturally significant: it enables AI model creators to own their work cryptographically. The Dobby model family on HuggingFace includes explicitly "unhinged" (uncensored) fine-tunes of Llama. This is ideologically aligned with sovereignty. But until GRID mainnet launches and proves the system works, it's architecture and aspiration, not operational freedom.

## Tokenomics

SENT has a maximum supply of 34,359,738,368 tokens (exactly 2^35; the computer science in the team shows). Distribution:

- **Community Initiatives & Airdrop:** 44% – 30% at TGE, 70% linear over 4 years. The largest single allocation.
- **Team:** 22% – 1-year cliff, linear over 6 years. The longest team vest in DeAI.
- **Ecosystem & R&D:** 19.55% – 30% at TGE, 70% linear over 4 years
- **Investors:** 12.45% – 1-year cliff, linear over 4 years
- **Public Sale:** 2% – 100% at TGE via Binance Wallet IEO at $0.01106

Additional 2% annual inflation directed to a Community Emission Pool. Unused amounts lock at the end of each year.

Circulating supply is approximately 21% of the maximum.

The next major event is the investor and team cliff unlock in November 2026, when 34.45% of supply (12.45% investors + 22% team) begins vesting. Given that the FDV is below the seed valuation, VCs have strong incentive to sell.

The FDV-to-market-cap ratio of 4.75x quantifies the dilution ahead. The Binance IEO was at $0.01106, so IEO buyers remain in profit at current levels. The listing price was $0.0379, so holders who bought at listing are currently underwater. Live price data is displayed above.

Staking is planned but not yet live. It will launch with GRID mainnet, which has no committed date. Staking will enable governance, reward eligibility, model access, and task qualification.

Listed on Binance, Coinbase, Kraken, Bybit, Gate.io, and KuCoin. All three major Western exchanges carry the token. For a pre-revenue project, this is unusually strong exchange coverage. Live price data is displayed above.

## How to participate

**Hold and wait.** Current token utility is limited to governance signalling. Staking isn't live. The per-invocation fee model isn't active. GRID mainnet (no committed date) is the dependency for meaningful participation. Technical skill: basic.

**Build on GRID.** Use ROMA framework and Sentient Agent Framework to build AI agents for The GRID. SDKs in Python and Rust. When GRID mainnet launches, agents earn per-invocation fees. Technical skill: advanced. Over $1M in grants and computing credits committed.

**Contribute models.** Fine-tune and deploy AI models with OML fingerprinting. Model creators will earn per-invocation revenue when their models are used on The GRID. Requires GPU hardware for fine-tuning and ML development skills. Technical skill: advanced.

## Honest assessment

## Freedom Score: 47/100

Sentient has strong open-source credentials and a novel approach to AI model sovereignty, but most systems are not yet decentralised.

**Infrastructure Decentralisation: 8/20.** GRID testnet runs agents across Base, Polygon, and Arbitrum. Sentient Enclaves Framework suggests TEE-based secure computation. However, GRID mainnet is not live (no committed date). Infrastructure depends on existing L2 chains. Currently centralised operation by Sentient Labs.

**Governance Decentralisation: 7/20.** Governance is planned but not yet active. The Sentient Foundation is a nonprofit, a structural commitment to community governance. Token-based governance is intended but mechanism not fully documented. All current decisions are made by the founding team.

**Token Distribution Fairness: 7/15.** The score leads with the position. On chain, about seven-tenths of supply sits in eight team or investor multisigs, across a holder base of under four thousand addresses, and none of the top ten is neutral infrastructure. The allocation reads better: 65.55% is community-facing, the team's 22% vests over six years and investors hold 12.45%. With only about a fifth of supply circulating, those community tokens have not reached a community yet.

**Censorship Resistance: 8/15.** OML enables model creators to distribute models openly while retaining cryptographic ownership, a novel censorship-resistance mechanism. Dobby Unhinged models are explicitly uncensored. ROMA is open-source. However, GRID infrastructure isn't yet decentralised. TEE enclaves depend on hardware vendors.

**Data Sovereignty: 8/15.** OML fingerprinting enables model ownership verification without centralised registries. Model creators retain control via cryptographic proofs. Sentient Enclaves use TEEs for private computation. The focus is model creator sovereignty; user data sovereignty isn't specifically addressed.

**Open Source Transparency: 9/15.** Seventeen non-fork repos carrying more than fourteen thousand stars, led by ROMA, OpenDeepSearch and OML Fingerprinting, with OML peer-reviewed and HuggingFace models published. The star count is real; what it is being read as does not survive a look at the licences. ROMA carries more than a third of those stars and has no licence file at all, checked against a repo in the same org that returns Apache-2.0 through the identical call, so the flagship of this dimension grants a reader no rights. It was last pushed in February, and ten of the seventeen have gone six months untouched, which makes the "all repos updated March 2026" line this review previously carried false rather than merely stale. There is also no published smart contract audit.

## Returns Score: 48/100

**Token Utility: 10/20.** SENT is planned for governance, staking, model access, and per-invocation fee distribution. However, staking is not live (GRID mainnet has no committed date). Current utility is limited to governance signalling and holding. The per-invocation fee model is designed but not yet implemented.

**Value Accrual: 10/20.** Per-invocation fee model will distribute revenue to model creators, deployers, and validators via smart contracts. This is a direct accrual mechanism: more usage equals more fee distribution. However, this mechanism isn't live. No revenue exists. Score reflects a credible but entirely forward-looking accrual path.

**Supply Dynamics: 8/20.** 34.36B max supply with only 21% circulating. Major investor and team cliff unlock in November 2026. 2% annual inflation is low. Team vesting is 6 years (positive). However, the FDV/market-cap ratio of 4.75x quantifies the dilution ahead. FDV trades below the seed valuation ($1.2B), suggesting VCs have incentive to sell at unlock.

**Revenue Sustainability: 7/25.** Zero on-chain revenue. GRID mainnet (no committed date) is the dependency. Per-invocation fee model is theoretically sound. Sentient claims a 2M+ Sentient Chat waitlist and 100K+ test queries, indicating prospective demand. But converting waitlist signups to paying users is unproven. $1M+ in grants committed.

**Liquidity & Access: 13/15.** Binance, Coinbase, Kraken, OKX and Bybit all quote SENT, as do Gate.io and KuCoin, across more than 30 venues in all, and that breadth keeps the score in the top band. Depth is the weak part: a 25 September CoinGecko read put daily volume well under the top band's threshold, with the largest share on a minor venue, GroveX. The tier-1 share also swings widely between reads, from under a tenth on 18 September to about a third on 24 September.  Turnover near 5% of market cap lets retail size trade without severe slippage but falls short of institutional depth, so the score sits at the bottom of the band.

### Quadrant: D (Low Freedom, Low Returns)

Sentient sits on the border of Quadrants B and D. The open-source commitment and community tokenomics point toward Quadrant B (sovereignty play), but the pre-mainnet state and zero revenue currently place it in Quadrant D.

### Key risks

- **GRID dependency.** The entire revenue model, staking mechanism, and meaningful token utility depend on GRID mainnet launching (no committed date). Delays would be material.
- **VC cliff unlock.** November 2026: 12.45% investor allocation begins vesting. Investors paid a $1.2B valuation and the market FDV sits below it. Strong sell incentive.
- **No audit.** No published smart contract audit for a project of this size, which Sentient's own MiCA whitepaper confirms by answering "no" to whether an audit was conducted. The token contract itself is verified: the SENT proxy `0x56a3ba04e95d34268a19b2a4474dc979babdaf76` was verified on 23 January 2026, with `SentientTokenV1` behind it.
- **Zero revenue.** Everything is forward-looking. The per-invocation model is sound in theory but unproven.
- **AGI expectations.** "Community-owned AGI" sets expectations that cannot realistically be met. Managing narrative vs reality will be an ongoing challenge.
- **21% circulating.** FDV/mcap ratio of 4.75x. Massive dilution ahead from multiple vesting schedules.
- **Competition.** Bittensor has a live network. Fetch-ASI has partnerships. Morpheus has active compute. Sentient has a waitlist.

## Score change log

| Date | Score | Change | Reason |
| - | - | - | - |
| 2026-10-07 | Data | N/A | Turnover restated from above 6% to near 5% of market cap (7-day median 4.8%, early October). Scores unchanged. |
| 2026-10-04 | Editorial | N/A | Product-in-practice sweep: the ROMA repo has had no commits since 16 February 2026 (GitHub API), now stated plainly. Research JSON drops the lapsed Q2 2026 GRID mainnet date. Scores unchanged. |
| 2026-10-04 | Freedom | Token Distribution Fairness 9 → 7 (Freedom 49 → 47, D held) | About seven-tenths of supply sits in team and investor multisigs, so the score now leads with the on-chain position rather than the community-facing allocation. |
| 2026-09-25 | Returns | Liquidity & Access 15 → 13 (Returns 50 → 48, D held) | Listed on all five majors, but daily volume sits well under the top band's threshold and the largest share trades on a minor venue, so the score drops to the bottom of the band. |
| 2026-09-24 | Editorial | N/A | Dropped the unverified "biggest VC bet in crypto-AI" superlative from the description and verdict; the $85M seed stands. Scores unchanged. |
| 2026-09-24 | Data | N/A | Removed the leftover "token contract not verified" line (corrected 5 Sept), the stale absolute FDV, and the 18 September venue split stated as current; a 24 September read put tier-1 venues at about a third of volume. Scores unchanged. |
| 2026-09-23 | Data | N/A | Returns summary aligned with the 6 September liquidity correction (tier-1 listings, thin volume on them). Scores unchanged. |
| 2026-09-23 | Data | N/A | Integrity records re-cut: 1 claim now quotes Sentient's own pages verbatim and 4 are our own measurements. Scores unchanged. |
| 2026-09-07 | Data | N/A | WS3b open-source rail: the star count is real and the openness reading does not survive the licences. ROMA holds over a third of the org's stars, has no licence file (controlled against a same-org repo returning Apache-2.0), and was last pushed in February, so "all repos updated March 2026" was false. Open Source 9/15 flagged for October. |
| 2026-09-06 | Data | N/A | WS3b distribution rail: on-chain top-10 verdict added. 69.64% of supply sits in eight team or investor multisigs (largest 15.00%) across 3,702 holders, with zero neutral infrastructure in the top ten. The allocation evidence describes the plan; this is where the tokens are. Token Distribution Fairness 9/15 flagged for October. |
| 2026-09-06 | Data | N/A | WS3b liquidity rail: all six tier-1 venues we name do quote SENT, but they carry 9.1% of reported volume while GroveX and Biconomy carry 55.3% between them, so the exceptional-tier-1-liquidity wording is corrected to the listings claim it supports. Liquidity & Access 15/15 flagged for October. |
| 2026-09-05 | Data | N/A | Fossilising-claim audit. "Token contract not verified on Etherscan" was wrong and is corrected: the SENT proxy was verified 23 Jan 2026, checked independently on Blockscout. The likely origin was reading a different, unverified SENT contract. Franklin Templeton is downgraded from investor to ecosystem participant, which is how Sentient's own announcement describes it. |
| 2026-08-19 | Data | N/A | Correction. The no-audit risk row priced the project at a hardcoded FDV, which drifts. Restated as project size; no score change. |
| 2026-08-15 | Data | N/A | Circulating supply de-hardcoded into the live `<CirculatingSupply />` component, percentage refreshed to 21.07% (CoinGecko). November 2026 cliff unlock framing unchanged. No score change. |
| 2026-08-10 | Data | No change | WS3 coldstart grounding, two claim-spine verdicts: supply dynamics `match` at onchain-reconciled (2^35 = 34.36B confirmed, no mint path, but the UUPS proxy leaves the cap mutable) and a token-layer censorship `established` anchor (the SENT ERC-20 is pausable and admin-upgradeable). Distribution deferred. Detail in the research JSON. |
| 2026-07-06 | Editorial | N/A | Added a "The product in practice" section (per `specs/product-comparison-framework.md`); the honest read is that Sentient is mostly vision (Dobby models and hosted chat are live, the GRID economic mainnet is not, inference is centralised, no protocol revenue yet). No score change. |
| 2026-06-26 | Data | N/A | DefiLlama LlamaAI harvest cross-check, verified independently (CoinGecko + api.llama.fi). Structure held: circulating 21.1% (~7.24B of 34.36B / 2^35), FDV/MCap 4.75x, ATH $0.048 (1 Feb 2026, now -73.9%). Refreshed the absolute FDV from $731M to ~$430M (CoinGecko, the token has fallen ~40% since the prior figure; the ratio and the below-seed-valuation thesis are unchanged and sharper). Corrected a launch-timeline conflation surfaced by the harvest: the TGE concluded 22 Nov 2025 (our date holds), but the Binance Wallet public sale was 19 Jan 2026 and spot trading opened 22 Jan 2026, so the prior "launched via Binance Wallet IEO" phrasing was tightened. No score change. |
| 2026-03-06 | Both | N/A | Initial publish. Freedom 49/100, Returns 50/100. |