Projects
Decentralised AI projects. Freedom scores, market data and honest assessment. Compare projects Top movers
| Project | Category | Freedom▼ | Returns | Quadrant | Status | Launch | Price | Mkt Cap | FDV | Trend | OYM Verdict | Review |
|---|---|---|---|---|---|---|---|---|---|---|---|---|
 Morpheus MOR | platform | 8 | 6 | A | active | fair launch | $1.99 | $15.3M | $15.8M | -10.5% -18.9% +61.9% -57.7% | One of the cleanest fair launches in DeAI: no insiders, no VC, a permissionless compute marketplace and Power Factor lock-ups that constrain selling. The weak spot is value capture. The buyback engine runs on staked-capital yield rather than compute demand, and it has shrunk with TVL while emissions still outpace burns. | Read |
 Flux FLUX | compute | 8 | 6 | A | active | mining | $0.044 | $18.2M | $18.2M | -14.6% -33.4% -55.7% -76.4% | One of crypto's fairest launches with verifiable infrastructure across 67 countries. Six years and ~100 deployed applications. The sovereignty credentials are excellent; the commercial traction is not. | Read |
 Golem Network GLM | compute | 7 | 5 | B | active | ico | $0.099 | $99.1M | $99.1M | -7.5% -25.2% -54.6% -54.5% | The OG decentralised compute network with one of crypto's fairest token distributions. Ten years of building with zero protocol revenue and a token that captures no value from usage. | Read |
 Akash Network AKT | compute | 7 | 7 | A | active | ico | $0.664 | $193.8M | $195.9M | -12.5% -19.1% +75.4% -39.1% | The most credible decentralised cloud marketplace operating today. Paying revenue, named customers, BME burn live since 23 March 2026. Provider count at network low; chain migration still looming. | Read |
| AntSeed ANTS | inference | 7 | 3 | B | active | fair launch | -- | -- | -- | -- | The architecturally honest version of what most DeAI inference marketplaces claim to be. BitTorrent DHT for discovery, WebRTC for transport, GPL-3.0 codebase, no pre-mine, no team allocation, no VC allocation. Reciprocal partnerships with Secret Network and Venice. Doxxed co-founder with NFTrade track record. 17 providers and verifiable on-chain settlements. ANTS transfers are intentionally disabled during the protocol's economic-architecture phase, which scores hard against Returns but isn't fatal: read this as Freedom 66/C with the Returns chapter not yet written. | Read |
 IoTeX IOTX | other | 6 | 5 | B | active | ico | $0.0028 | $26.2M | $26.2M | -10.2% -32.6% -60.5% -86.7% | The most technically differentiated DePIN chain with verifiable innovations in device identity and ZK verification. Seven years of building with ~$1.2M TVL, but MiCA compliance, CF Benchmarks listing, and the Real-World AI Foundry signal institutional maturation. The ioTube bridge exploit ($4.4M, February 2026) is a setback. | Read |
 NEAR Protocol NEAR | platform | 6 | 7 | A | active | ico | $1.91 | $2.48B | $2.48B | -11.4% -24.0% +19.7% -14.2% | The strongest AI credentials of any L1 founder, a proven sharded chain running since 2020, and real cross-chain traction via Intents ($19B+ cumulative volume). Fee switch and inflation halving strengthen tokenomics. But TVL remains thin, AI products are pre-traction, and the inflation halving was pushed through despite a failed governance vote. | Read |
 Olas OLAS | agent | 6 | 4 | B | active | ico | $0.030 | $6.7M | $14.1M | -13.6% +0.6% -66.6% -88.0% | Measurable on-chain agent usage and strong open-source credentials, but the economics are not working. $89K marketplace turnover, 99.6% token decline, and 42.65% insider allocation with no enforced vesting. A technically credible protocol that has not yet solved its tokenomics. | Read |
 Oasis Network ROSE | data | 6 | 5 | B | active | ico | $0.0059 | $46.0M | $59.1M | -11.6% -35.4% -44.7% -74.4% | The technology works and is finally shipping product. ROFL mainnet, institutional pilots, liquid staking. But TVL peaked at $7.5M (May 2025) then fell back sharply. Promising direction, unproven traction. | Read |
 OriginTrail TRAC | other | 6 | 7 | A | active | ico | $0.277 | $124.0M | $138.6M | -19.3% -29.6% -27.2% -14.0% | The quiet achiever of DeAI. Best tokenomics, named enterprise adoption, unique category. Eight years of building with unquantified revenue. | Read |
 Theta Network THETA | compute | 6 | 5 | B | active | ico | $0.131 | $131.3M | $131.3M | -16.5% -28.3% -50.8% -79.8% | Enterprise-backed infrastructure with Google, Samsung, and Sony as validators. Impressive partnerships but THETA's value accrual problem is fundamental: all fees flow to TFUEL, not THETA. | Read |
 Vana VANA | data | 6 | 6 | A | active | airdrop | $1.07 | $32.8M | $127.9M | -7.1% -23.0% -57.0% -75.4% | The most technically complete data sovereignty protocol in crypto, with reported user adoption and strong backing. But permissioned validators, Foundation-controlled treasury, and a ~97% price decline tell you the market is not yet convinced that 'own your data' translates to token value. | Read |
 Nosana NOS | compute | 6 | 5 | B | active | ico | $0.279 | $27.9M | $27.9M | -0.4% -33.9% +33.4% -54.6% | Working inference marketplace with named partner deployments and Render integration. Undisclosed revenue and thin liquidity make it a conviction bet on the inference thesis. | Read |
 Venice VVV | inference | 6 | 7 | A | active | airdrop | $13.43 | $631.5M | $1.08B | -6.6% -15.7% +699.8% +392.0% | Best privacy-focused inference platform available. Now with verifiable E2EE and TEE modes via NEAR and Phala. Centralised company but aligned incentives. | Read |
 Bittensor TAO | platform | 6 | 6 | A | active | mining | $210.51 | $2.02B | $4.42B | -8.8% -15.1% -3.9% -34.9% | Largest DeAI network with live workloads. PoA block production and governance far more centralised than marketed. Covenant AI's April 2026 exit alleges specific governance abuses now under active review. No security audit. | Read |
 Nillion NIL | data | 6 | 4 | B | active | ico | $0.034 | $16.5M | $33.9M | -14.8% -47.0% -54.0% -87.8% | The right thesis (AI needs privacy-preserving computation) but battered by a market maker scandal and zero commercial validation. Exceptional team, credible technology. Blacklight verification layer is a real deliverable, but revenue remains absent. | Read |
| Openmind ROBO | agent | 6 | 4 | B | active | ico | $0.017 | $37.1M | $166.2M | -14.2% -13.3% -59.2% -59.2% | Openmind ships a working open-source robotics runtime with major OEM partners and a doxxed Stanford / MIT CSAIL founding team. The token thesis is materially weaker. ROBO is the work-bond and emission token; USDC settles the actual machine-to-machine payments. A 44.3% insider cliff begins linear vesting in February 2027. | Read |
 FLock.io FLOCK | compute | 6 | 6 | A | active | airdrop | $0.034 | $13.7M | $33.5M | -13.0% -38.9% -64.0% -79.0% | NeurIPS-awarded federated learning meets blockchain. $2.7M in protocol revenue from a small number of tasks. The academic credentials hold up; the question is whether 16 training tasks in 10 months becomes 1,600. | Read |
 Ocean Protocol OCEAN | data | 6 | 4 | B | active | ico | $0.116 | $28.0M | $31.2M | +13.7% -3.9% -26.5% -59.2% | The technology substantially outperforms the tokenomics and governance situation. Compute-to-Data remains one of the most elegant solutions to the data privacy problem in Web3, but a token without stated utility, a fragmented community after 81% supply conversion, dismantled governance, and a settled lawsuit returning 286M FET make this a project to watch technically and avoid as a token investment. | Read |
| Phala Network PHA | compute | 6 | 5 | B | active | ico | $0.031 | $26.0M | $30.9M | -16.7% -22.2% -10.7% -66.8% | Confidential computing tech with enterprise compliance credentials. The privacy works. The decentralisation does not, yet. | Read |
 peaq PEAQ | other | 5 | 5 | D | active | ico | $0.021 | $47.9M | $93.4M | -6.1% -33.9% -35.4% -73.3% | A well-engineered Machine Economy L1 with 60+ DePINs and enterprise partnerships, but $163.6K annualised revenue tells you the market is not paying for it yet. | Read |
 Allora Network ALLO | other | 5 | 5 | D | active | ico | $0.302 | $71.1M | $238.1M | -23.9% +19.9% +180.9% -33.3% | Notable technical innovation with novel inference synthesis and tier-1 exchange listings from day one. Unproven revenue model and a massive vesting cliff in November 2026 that could lift circulating supply by roughly two-thirds. | Read |
 ElizaOS ELIZAOS | agent | 5 | 3 | D | active | fair launch | $0.0006 | $4.5M | $5.2M | -15.0% -37.9% -90.6% -94.1% | A well-built open-source AI agent framework whose founding company (Eliza Labs) is winding down amid a federal class action. The framework has daily commits and v3 in beta; the token is near zero with no value capture mechanism. Open-source users are fine. Tokenholders are not. | Read |
 Gensyn AI | compute | 5 | 3 | D | active | ico | $0.022 | $28.8M | $220.5M | -17.3% -33.1% -61.2% -61.2% | Serious ML research and strong open-source credentials, with a token that finally trades after a 29 April 2026 TGE. The training-network mainnet is still unfinished, governance is team-controlled, and 54.6% of supply sits with insiders. A credible research bet now with a live market price attached. | Read |
| Nous Research Pre-TGE | compute | 5 | 2 | D | upcoming | pre token | -- | -- | -- | -- | The strongest decentralised pre-training story in DeAI. Consilience 40B trained over the internet for 20 trillion tokens on Psyche. Hermes 4.3 trained start-to-finish on the same network. The catch sits in the cap table. A $1B SAFT round led by Paradigm puts venture capital between the technology and any future public token holder, and TGE terms are still unannounced. | Read |
 Fetch.ai / ASI Alliance FET | agent | 5 | 6 | C | active | ico | $0.174 | $393.5M | $472.9M | -6.5% -27.3% -14.2% -73.5% | Functional agent runtime and open-source foundation, but the merger has been messy, usage metrics are opaque, and core products are centrally operated. | Read |
 Auki AUKI | other | 5 | 4 | D | active | ico | $0.0050 | $22.0M | $49.6M | -12.3% -38.3% -44.3% -63.7% | Auki ships a working retail spatial-AI product with one externally documented pilot and an honest, OpenZeppelin-acknowledged centralisation posture. The tokenomics design is unusually disciplined for a 2024 launch. The economic flywheel is trivial today: $166,400 of trailing-year on-chain fees, deeply negative net of incentives. The 40,000-stores figure that surfaced in DeAI commentary is unsourced and most likely a confusion with an unrelated retail-closure headline. | Read |
 Sentient SENT | inference | 5 | 5 | D | active | ico | $0.013 | $92.8M | $440.3M | -17.6% -14.2% -56.4% -56.4% | The biggest VC bet in crypto-AI: $85M seed from Founders Fund and Pantera, OML is a genuine innovation. But GRID mainnet is not live, there is zero revenue, and November 2026 VC unlock looms. | Read |
 Ora Protocol ORA | other | 5 | 4 | D | active | fair launch | $0.012 | $648K | $3.8M | +32.9% -47.9% -71.7% -97.7% | Genuine technical innovation in verifiable AI inference with ERC standards contributions. But the token is down 99.85%, effectively illiquid, and 83% of supply is locked with undisclosed allocation. | Read |
 NuNet NTX | compute | 5 | 3 | D | active | ico | $0.0021 | $1.8M | $2.1M | -9.0% -22.5% -69.6% -79.5% | The architecturally honest end of the DeAI compute shelf. Apache 2.0 across the orchestration stack, multi-chain settlement on Ethereum, Cardano and BSC, and a mainnet that publishes its actual device count instead of an inflated one. The cost of that honesty is a $3M market cap, an audit four years and eight months old, foundation-only governance, and a token whose design accommodates its own substitution by stablecoins. | Read |
 Warden Protocol WARD | agent | 5 | 5 | D | active | ico | $0.0078 | $2.0M | $7.8M | -20.1% -27.8% -91.0% -91.0% | An AI agent L1 from Qredo/Binance alumni. On-chain activity checks out (~35M+ txs verified via RPC sampling), but the 60M agent runs headline includes off-chain counts we can't verify. Inflated user numbers, proprietary core modules, 42% insider allocation, and a sustained ~94% decline since the Feb 2026 peak demand caution. | Read |
| Giza GIZA | agent | 4 | 3 | D | active | ico | $0.0037 | $1.3M | $3.7M | -7.2% -35.3% -92.3% -98.2% | Retired its two shipping agents in March 2026 and returned user funds, then relaunched as a single unified agent whose headline activity is a stale snapshot with no verifiable current AUA or fees. The ZKML pillar that brands the project is still roadmap and rests on a single committer. | Read |
 Intelligent Internet FC | platform | 4 | 2 | D | upcoming | pre token | -- | -- | -- | -- | Strong open-source AI output (9 models, 20 datasets on HuggingFace) and a functional coding agent, but no chain, no token, and execution depends on a controversial founder. Watch closely. | Read |
 Virtuals Protocol VIRTUAL | agent | 4 | 7 | C | active | ico | $0.538 | $353.3M | $537.7M | -10.4% -25.5% -21.8% -63.6% | The dominant AI agent launchpad, now spanning Base, Solana, Ronin, Arbitrum, and XRP Ledger, with on-chain revenue and a buyback-and-burn programme. Centralised inference infrastructure and pump-and-dump agent token dynamics persist, but the Unicorn launch system and deflationary mechanisms show a platform maturing beyond its pump.fun origins. | Read |
 io.net IO | compute | 4 | 5 | D | active | ico | $0.176 | $61.4M | $140.5M | -5.2% +3.0% +19.1% -74.0% | The IDE burn went live on-chain in June 2026, but io.net's own explorer shows it is emission-funded rather than the revenue buyback marketed, with demand-driven emissions still off. Revenue is ~$12.5M annualised, below the $20M+ once implied. Closed-source core, no governance, and inflated GPU metrics persist. A centralised GPU marketplace with a token, not a decentralised protocol. | Read |
| Heurist HEU | compute | 4 | 3 | D | active | airdrop | $0.0029 | $581K | $2.9M | +0.3% -47.1% -76.6% -85.9% | A functional open-source agent infrastructure stack attached to a paused DePIN mining programme. Strong GitHub traction (812 stars on the agent SDK), doxxed New York team, multi-chain HEU contracts. But the mining season has been paused since 18 January 2025, the agent payment layer settles in USDC rather than HEU, and the token is 98% below its TGE-week all-time high on persistent insider vesting against a 20% float. The product is underwriteable; the decentralisation narrative is currently aspirational. | Read |
 Grass GRASS | data | 4 | 6 | C | active | airdrop | $0.506 | $308.7M | $506.3M | +23.5% +7.4% +75.5% -56.4% | Impressive scale and reported enterprise traction, but this is a centralised data business with a token incentive layer. Code is entirely closed-source, governance exists only in documentation, users contribute bandwidth without any sovereignty over the data produced, and all revenue flows through an opaque Foundation. The widest airdrop in Solana history doesn't make a network decentralised. | Read |
 Sahara AI SAHARA | data | 3 | 5 | D | active | ico | $0.012 | $40.4M | $115.6M | -11.2% -65.4% -55.6% -86.5% | Strong academic credentials, shipping data labelling traction, and top-tier VC backing. But the core blockchain is not operational, every decentralisation feature is aspirational, and the token contract includes a pause function. A well-funded centralised AI platform with blockchain aspirations, not a decentralised network. | Read |
 Render Network RENDER | compute | 3 | 7 | C | active | ico | $1.58 | $822.3M | $845.7M | -7.2% -18.6% +23.3% -48.8% | Hollywood rendering demand with paying studio customers. But permissioned network, proprietary core, and OTOY controls everything. | Read |
 Aethir ATH | compute | 3 | 6 | C | active | ico | $0.0042 | $85.3M | $177.9M | -9.6% -27.6% -53.4% -84.9% | The best revenue story in DeAI, attached to the worst decentralisation credentials. Governance is finally arriving, but this remains an enterprise GPU business with a token. | Read |
 OpenServ SERV | agent | 2 | 3 | D | active | ico | $0.036 | $27.4M | $35.6M | -29.9% -44.2% +152.9% -20.0% | Honest operator, credible research, immature implementation. The BRAID preprint and a credentialed external co-author check out, and the TypeScript SDK is properly open. Everything around them overstates: 122x marketing against a 74x paper, '400+ services' against about 31, an 'outside adoption' story that is roughly $1.3M of launchpad micro-caps, a flagship enterprise proof nobody can locate, and a buyback-burn that has never run on-chain. No security audit for a token live on two chains. Read this as Freedom 24/F and Returns 34/F, with the research the only part that earns credit. | Read |
 Cookie DAO COOKIE | data | 2 | 5 | D | active | ico | $0.0080 | $6.2M | $8.0M | -12.2% -33.7% -79.2% -95.1% | A legitimate AI agent analytics product with an industry-cited metric and first-mover advantage. But the 'DAO' label is marketing: zero open source code, no verifiable governance, fully centralised infrastructure. | Read |
Market data via CoinGecko. Updated every 6 hours.
Morpheus MOR One of the cleanest fair launches in DeAI: no insiders, no VC, a permissionless compute marketplace and Power Factor lock-ups that constrain selling. The weak spot is value capture. The buyback engine runs on staked-capital yield rather than compute demand, and it has shrunk with TVL while emissions still outpace burns.
Flux FLUX One of crypto's fairest launches with verifiable infrastructure across 67 countries. Six years and ~100 deployed applications. The sovereignty credentials are excellent; the commercial traction is not.
Golem Network GLM The OG decentralised compute network with one of crypto's fairest token distributions. Ten years of building with zero protocol revenue and a token that captures no value from usage.
Akash Network AKT The most credible decentralised cloud marketplace operating today. Paying revenue, named customers, BME burn live since 23 March 2026. Provider count at network low; chain migration still looming.
The architecturally honest version of what most DeAI inference marketplaces claim to be. BitTorrent DHT for discovery, WebRTC for transport, GPL-3.0 codebase, no pre-mine, no team allocation, no VC allocation. Reciprocal partnerships with Secret Network and Venice. Doxxed co-founder with NFTrade track record. 17 providers and verifiable on-chain settlements. ANTS transfers are intentionally disabled during the protocol's economic-architecture phase, which scores hard against Returns but isn't fatal: read this as Freedom 66/C with the Returns chapter not yet written.
IoTeX IOTX The most technically differentiated DePIN chain with verifiable innovations in device identity and ZK verification. Seven years of building with ~$1.2M TVL, but MiCA compliance, CF Benchmarks listing, and the Real-World AI Foundry signal institutional maturation. The ioTube bridge exploit ($4.4M, February 2026) is a setback.
NEAR Protocol NEAR The strongest AI credentials of any L1 founder, a proven sharded chain running since 2020, and real cross-chain traction via Intents ($19B+ cumulative volume). Fee switch and inflation halving strengthen tokenomics. But TVL remains thin, AI products are pre-traction, and the inflation halving was pushed through despite a failed governance vote.
Olas OLAS Measurable on-chain agent usage and strong open-source credentials, but the economics are not working. $89K marketplace turnover, 99.6% token decline, and 42.65% insider allocation with no enforced vesting. A technically credible protocol that has not yet solved its tokenomics.
Oasis Network ROSE The technology works and is finally shipping product. ROFL mainnet, institutional pilots, liquid staking. But TVL peaked at $7.5M (May 2025) then fell back sharply. Promising direction, unproven traction.
OriginTrail TRAC The quiet achiever of DeAI. Best tokenomics, named enterprise adoption, unique category. Eight years of building with unquantified revenue.
Theta Network THETA Enterprise-backed infrastructure with Google, Samsung, and Sony as validators. Impressive partnerships but THETA's value accrual problem is fundamental: all fees flow to TFUEL, not THETA.
Vana VANA The most technically complete data sovereignty protocol in crypto, with reported user adoption and strong backing. But permissioned validators, Foundation-controlled treasury, and a ~97% price decline tell you the market is not yet convinced that 'own your data' translates to token value.
Nosana NOS Working inference marketplace with named partner deployments and Render integration. Undisclosed revenue and thin liquidity make it a conviction bet on the inference thesis.
Venice VVV Best privacy-focused inference platform available. Now with verifiable E2EE and TEE modes via NEAR and Phala. Centralised company but aligned incentives.
Bittensor TAO Largest DeAI network with live workloads. PoA block production and governance far more centralised than marketed. Covenant AI's April 2026 exit alleges specific governance abuses now under active review. No security audit.
Nillion NIL The right thesis (AI needs privacy-preserving computation) but battered by a market maker scandal and zero commercial validation. Exceptional team, credible technology. Blacklight verification layer is a real deliverable, but revenue remains absent.
Openmind ships a working open-source robotics runtime with major OEM partners and a doxxed Stanford / MIT CSAIL founding team. The token thesis is materially weaker. ROBO is the work-bond and emission token; USDC settles the actual machine-to-machine payments. A 44.3% insider cliff begins linear vesting in February 2027.
FLock.io FLOCK NeurIPS-awarded federated learning meets blockchain. $2.7M in protocol revenue from a small number of tasks. The academic credentials hold up; the question is whether 16 training tasks in 10 months becomes 1,600.
Ocean Protocol OCEAN The technology substantially outperforms the tokenomics and governance situation. Compute-to-Data remains one of the most elegant solutions to the data privacy problem in Web3, but a token without stated utility, a fragmented community after 81% supply conversion, dismantled governance, and a settled lawsuit returning 286M FET make this a project to watch technically and avoid as a token investment.
Confidential computing tech with enterprise compliance credentials. The privacy works. The decentralisation does not, yet.
peaq PEAQ A well-engineered Machine Economy L1 with 60+ DePINs and enterprise partnerships, but $163.6K annualised revenue tells you the market is not paying for it yet.
Allora Network ALLO Notable technical innovation with novel inference synthesis and tier-1 exchange listings from day one. Unproven revenue model and a massive vesting cliff in November 2026 that could lift circulating supply by roughly two-thirds.
ElizaOS ELIZAOS A well-built open-source AI agent framework whose founding company (Eliza Labs) is winding down amid a federal class action. The framework has daily commits and v3 in beta; the token is near zero with no value capture mechanism. Open-source users are fine. Tokenholders are not.
Gensyn AI Serious ML research and strong open-source credentials, with a token that finally trades after a 29 April 2026 TGE. The training-network mainnet is still unfinished, governance is team-controlled, and 54.6% of supply sits with insiders. A credible research bet now with a live market price attached.
The strongest decentralised pre-training story in DeAI. Consilience 40B trained over the internet for 20 trillion tokens on Psyche. Hermes 4.3 trained start-to-finish on the same network. The catch sits in the cap table. A $1B SAFT round led by Paradigm puts venture capital between the technology and any future public token holder, and TGE terms are still unannounced.
Fetch.ai / ASI Alliance FET Functional agent runtime and open-source foundation, but the merger has been messy, usage metrics are opaque, and core products are centrally operated.
Auki AUKI Auki ships a working retail spatial-AI product with one externally documented pilot and an honest, OpenZeppelin-acknowledged centralisation posture. The tokenomics design is unusually disciplined for a 2024 launch. The economic flywheel is trivial today: $166,400 of trailing-year on-chain fees, deeply negative net of incentives. The 40,000-stores figure that surfaced in DeAI commentary is unsourced and most likely a confusion with an unrelated retail-closure headline.
Sentient SENT The biggest VC bet in crypto-AI: $85M seed from Founders Fund and Pantera, OML is a genuine innovation. But GRID mainnet is not live, there is zero revenue, and November 2026 VC unlock looms.
Ora Protocol ORA Genuine technical innovation in verifiable AI inference with ERC standards contributions. But the token is down 99.85%, effectively illiquid, and 83% of supply is locked with undisclosed allocation.
NuNet NTX The architecturally honest end of the DeAI compute shelf. Apache 2.0 across the orchestration stack, multi-chain settlement on Ethereum, Cardano and BSC, and a mainnet that publishes its actual device count instead of an inflated one. The cost of that honesty is a $3M market cap, an audit four years and eight months old, foundation-only governance, and a token whose design accommodates its own substitution by stablecoins.
Warden Protocol WARD An AI agent L1 from Qredo/Binance alumni. On-chain activity checks out (~35M+ txs verified via RPC sampling), but the 60M agent runs headline includes off-chain counts we can't verify. Inflated user numbers, proprietary core modules, 42% insider allocation, and a sustained ~94% decline since the Feb 2026 peak demand caution.
Retired its two shipping agents in March 2026 and returned user funds, then relaunched as a single unified agent whose headline activity is a stale snapshot with no verifiable current AUA or fees. The ZKML pillar that brands the project is still roadmap and rests on a single committer.
Intelligent Internet FC Strong open-source AI output (9 models, 20 datasets on HuggingFace) and a functional coding agent, but no chain, no token, and execution depends on a controversial founder. Watch closely.
Virtuals Protocol VIRTUAL The dominant AI agent launchpad, now spanning Base, Solana, Ronin, Arbitrum, and XRP Ledger, with on-chain revenue and a buyback-and-burn programme. Centralised inference infrastructure and pump-and-dump agent token dynamics persist, but the Unicorn launch system and deflationary mechanisms show a platform maturing beyond its pump.fun origins.
io.net IO The IDE burn went live on-chain in June 2026, but io.net's own explorer shows it is emission-funded rather than the revenue buyback marketed, with demand-driven emissions still off. Revenue is ~$12.5M annualised, below the $20M+ once implied. Closed-source core, no governance, and inflated GPU metrics persist. A centralised GPU marketplace with a token, not a decentralised protocol.
A functional open-source agent infrastructure stack attached to a paused DePIN mining programme. Strong GitHub traction (812 stars on the agent SDK), doxxed New York team, multi-chain HEU contracts. But the mining season has been paused since 18 January 2025, the agent payment layer settles in USDC rather than HEU, and the token is 98% below its TGE-week all-time high on persistent insider vesting against a 20% float. The product is underwriteable; the decentralisation narrative is currently aspirational.
Grass GRASS Impressive scale and reported enterprise traction, but this is a centralised data business with a token incentive layer. Code is entirely closed-source, governance exists only in documentation, users contribute bandwidth without any sovereignty over the data produced, and all revenue flows through an opaque Foundation. The widest airdrop in Solana history doesn't make a network decentralised.
Sahara AI SAHARA Strong academic credentials, shipping data labelling traction, and top-tier VC backing. But the core blockchain is not operational, every decentralisation feature is aspirational, and the token contract includes a pause function. A well-funded centralised AI platform with blockchain aspirations, not a decentralised network.
Render Network RENDER Hollywood rendering demand with paying studio customers. But permissioned network, proprietary core, and OTOY controls everything.
Aethir ATH The best revenue story in DeAI, attached to the worst decentralisation credentials. Governance is finally arriving, but this remains an enterprise GPU business with a token.
OpenServ SERV Honest operator, credible research, immature implementation. The BRAID preprint and a credentialed external co-author check out, and the TypeScript SDK is properly open. Everything around them overstates: 122x marketing against a 74x paper, '400+ services' against about 31, an 'outside adoption' story that is roughly $1.3M of launchpad micro-caps, a flagship enterprise proof nobody can locate, and a buyback-burn that has never run on-chain. No security audit for a token live on two chains. Read this as Freedom 24/F and Returns 34/F, with the research the only part that earns credit.
Cookie DAO COOKIE A legitimate AI agent analytics product with an industry-cited metric and first-mover advantage. But the 'DAO' label is marketing: zero open source code, no verifiable governance, fully centralised infrastructure.