What is the difference between Celar and Zama?

Zama is a cross-chain confidentiality layer: an FHE coprocessor network that sits on top of existing chains, with a 13-node threshold KMS running inside AWS Nitro Enclaves. Celar is a sovereign Layer 1 where threshold FHE is the protocol's own data plane, secured by a 100-seat committee with a threshold decryption quorum, no TEE dependency, and slashable, attributable partial decryptions.

What is the difference between FHE chains and ZK chains like Aztec or Aleo?

ZK lets a user prove statements about their own data; shared state stays public. FHE lets the chain compute directly on encrypted shared state nobody can read — private orderbooks, sealed-bid auctions, encrypted balances. Celar combines both: protocol-level threshold FHE for shared encrypted state, plus a ZK shielded pool whose privacy holds even if every validator colludes.

Celar is in Phase 1 engineering against the v0.9.26 specification. The docs and a falsifiable protocol-health metric are published at celar.network; the full technical specification is shared under NDA in the data room.

Celar vs Zama vs Fhenix vs Inco vs Aztec vs Aleo

01 /

The matrix

AS OF JULY 2026 · SOURCES BELOW · CORRECTIONS: PRESS@CELAR.NETWORK

	CELAR	ZAMA	FHENIX	INCO	AZTEC	ALEO
ARCHITECTURE	Sovereign Layer 1privacy is the protocol's data plane	Cross-chain confidentiality layer¹coprocessor atop host chains	FHE coprocessor (CoFHE)²pivoted from L2	Confidentiality layer³on Base, Solana	ZK rollup L2 on Ethereum⁴	ZK Layer 1⁵
PRIVACY PRIMITIVE	Threshold FHE + stealth addresses + ZK shielded pool	TFHE (coprocessor)	FHE (coprocessor)	TEE-based encrypted compute³	ZK proofs, client-side proving	ZK proofs (snarkVM)
SHARED ENCRYPTED STATE	YES — protocol levelprivate orderbooks, sealed bids, encrypted balances	YESvia coprocessor on host chain	YESvia coprocessor	HARDWARE-TRUSTEDinside enclaves	NOpublic functions are plaintext	NOown-data proofs
DECRYPTION KEY CUSTODY	Threshold committee (100 seats)24-of-100 confidentiality as deployed · 79-of-100 design target (whitepaper §7.1) · epoch share refresh · signed, attributable, slashable partials · no TEE dependency	13-node KMS, 2/3 ruleMPC inside AWS Nitro Enclaves¹	Threshold networkparameters unpublished²	TEE vendor trusthardware root	None neededuser-local keys — a real ZK strength	None neededuser-local keys
PRIVACY IF ALL OPERATORS COLLUDE	HOLDSZK shielded pool tier survives full committee collusion	BREAKS≥2/3 of 13 KMS nodes decrypt	BREAKSthreshold operators	BREAKScompromised enclave	HOLDS	HOLDS
EVM / SOLIDITY	EVM-EQUIVALENTzero new opcodes; encrypted types via precompiles	SOLIDITYon host EVM chains	SOLIDITY	SOLIDITY	NOIR DSLnew language, new tooling	LEO DSLnew language
MEV	NONE BY CONSTRUCTIONcontent-blind randomized ordering, no public mempool	INHERITS HOST CHAIN	INHERITS HOST CHAIN	INHERITS HOST CHAIN	SEQUENCER POLICY	VALIDATOR POLICY
PUBLISHED LEAKAGE MATRIX	YESwhat leaks, to whom, with mitigation roadmap	NO	NO	NO	NO	NO
FALSIFIABLE SECURITY METRIC	YES — ℋ ≥ 2live, recomputable by anyone: celar.network/health	NO	NO	NO	NO	NO
LEAK-REJECTING COMPILER	YEScode that would leak encrypted data fails the build	NO	NO	NO	TYPE-LEVEL PRIVACY	TYPE-LEVEL PRIVACY
STATUS — JULY 2026	PHASE 1 ENGINEERINGspec v0.9.26 · 8 review rounds, 34 findings closed	LIVEEthereum mainnet¹	TESTNETSSepolia, Arb, Base²	LIVEBase, Solana³	LIVEIgnition, Nov 2025⁴	LIVEsince Sept 2024⁵

02 /

How to read this honestly

Every project on this page is credible, and two rows above are wins for our competitors: Zama, Inco, Aztec, and Aleo are live while Celar is in Phase 1 engineering — and ZK chains need no global decryption key at all. We publish the matrix anyway, because the remaining rows are structural: they don't change with time or funding.

THE FHE COPROCESSORS — ZAMA · FHENIX · INCORented privacy inherits its landlord

A confidentiality layer on top of Ethereum or Base inherits the host chain's mempool, its MEV economy, its gas market, and its ordering policy. And custody concentrates: Zama's key lives with 13 KMS nodes running inside AWS Nitro Enclaves — a hardware and cloud-vendor trust root. Celar's committee is 100 seats, with confidentiality against any 24 of 100 as deployed (79-of-100 the design target — whitepaper §7.1): 24 learn nothing, and reconstruction requires strictly more. No enclave is required, and every partial decryption is signed and slashable at protocol level.

THE ZK CHAINS — AZTEC · ALEOProving your data isn't computing on it

ZK is the right tool for own-data privacy, and Aztec's client-side proving is excellent work. But ZK alone cannot run a private orderbook or a sealed-bid auction — shared state someone computes on but nobody reads. That requires FHE. Celar ships both primitives: FHE for shared encrypted state, and a ZK shielded pool for the trustless tier — so the "if everyone colludes" row holds without giving up encrypted computation.

CELARThe category is the intersection

Sovereign L1 economics (attack cost priced in staked $CELAR, measured live as ℋ), protocol-level threshold FHE with no TEE in the trust root, EVM equivalence with zero new opcodes, MEV removed by construction — and the honesty infrastructure: a leakage matrix and a falsifiable health metric no other chain publishes. That intersection is empty today. Celar is being built to fill it.

Read the Docs FHE vs ZK, in depth

03 /

Sources

[1] Zama Protocol litepaper — "not a new L1 or L2, but a cross-chain confidentiality layer"; 13-node KMS, 2/3 majority, MPC inside AWS Nitro Enclaves; Ethereum mainnet live. docs.zama.org

[2] Fhenix — CoFHE coprocessor for EVM chains; live on Ethereum/Arbitrum/Base Sepolia testnets; threshold decryption network. fhenix.io

[3] Inco — Lightning launched on Base and Solana; TEE-based encrypted compute. inco.org

[4] Aztec — privacy-preserving ZK L2 on Ethereum; Noir contracts; client-side proving; Ignition chain live Nov 2025. aztec.network

[5] Aleo — ZK L1, snarkVM, Leo language; mainnet since September 2024. aleo.org

Comparisons reflect public documentation as of July 2026. Projects evolve — if a row is out of date, email press@celar.network and we will fix it. That is what falsifiable means.