Post-Quantum Key and Signature Sizes: ML-KEM, ML-DSA, SLH-DSA, Falcon vs RSA and ECC (2026 Reference)
Byte sizes for every ML-KEM, ML-DSA, SLH-DSA and Falcon parameter set against RSA and ECC, each cited to its FIPS or RFC section.
An ML-DSA-65 signature is 3,309 bytes. An ECDSA P-256 signature is 64. That jump of more than 50 times, set out in FIPS 204, is the number that drives most post-quantum migration cost in TLS, certificates and tokens. This page lists the byte sizes of every NIST post-quantum parameter set against RSA and elliptic-curve cryptography (ECC), with the standard and section each figure comes from.
All sizes are raw byte lengths as the standards define them, without ASN.1, JSON or protocol framing. For the migration plan that uses these numbers, see the enterprise PQC migration guide for authentication.
ML-KEM key and ciphertext sizes (FIPS 203)
ML-KEM (Module-Lattice-Based Key-Encapsulation Mechanism) replaces RSA and Diffie-Hellman key exchange. Source: FIPS 203, Section 8, Table 3 (sizes) and the security categories stated in the same section.
| Parameter set | Encapsulation (public) key | Decapsulation (private) key | Ciphertext | Shared secret | NIST category |
|---|---|---|---|---|---|
| ML-KEM-512 | 800 | 1,632 | 768 | 32 | 1 |
| ML-KEM-768 | 1,184 | 2,400 | 1,088 | 32 | 3 |
| ML-KEM-1024 | 1,568 | 3,168 | 1,568 | 32 | 5 |
Bytes. Source: NIST FIPS 203, Section 8, Table 3.
ML-DSA key and signature sizes (FIPS 204)
ML-DSA (Module-Lattice-Based Digital Signature Algorithm) is the general-purpose replacement for RSA and ECDSA signatures. Source: FIPS 204, Section 4, Table 2 (sizes) and Table 1 (claimed security category).
| Parameter set | Public key | Private key | Signature | NIST category |
|---|---|---|---|---|
| ML-DSA-44 | 1,312 | 2,560 | 2,420 | 2 |
| ML-DSA-65 | 1,952 | 4,032 | 3,309 | 3 |
| ML-DSA-87 | 2,592 | 4,896 | 4,627 | 5 |
Bytes. Source: NIST FIPS 204, Section 4, Tables 1 and 2. FIPS 204 notes the private key can instead be stored as its 32-byte seed; RFC 9881 (Appendix B) gives that 32-byte seed form for X.509. NIST IR 8547 adds that ML-DSA-44 drops to category 1 if the random bit generator offers less than 192 bits of security.
SLH-DSA key and signature sizes (FIPS 205)
SLH-DSA (Stateless Hash-Based Digital Signature Algorithm) rests only on hash-function security. FIPS 205 approves 12 parameter sets. The SHA2 and SHAKE variant of each size share identical byte lengths; "s" sets favour small signatures, "f" sets favour fast signing. Source: FIPS 205, Section 11, Table 2 (public key, signature, category). Private keys are 4n bytes, twice the 2n-byte public key, per Section 3.1.
| Parameter sets | n | Public key | Private key | Signature | NIST category |
|---|---|---|---|---|---|
| SLH-DSA-SHA2-128s, SLH-DSA-SHAKE-128s | 16 | 32 | 64 | 7,856 | 1 |
| SLH-DSA-SHA2-128f, SLH-DSA-SHAKE-128f | 16 | 32 | 64 | 17,088 | 1 |
| SLH-DSA-SHA2-192s, SLH-DSA-SHAKE-192s | 24 | 48 | 96 | 16,224 | 3 |
| SLH-DSA-SHA2-192f, SLH-DSA-SHAKE-192f | 24 | 48 | 96 | 35,664 | 3 |
| SLH-DSA-SHA2-256s, SLH-DSA-SHAKE-256s | 32 | 64 | 128 | 29,792 | 5 |
| SLH-DSA-SHA2-256f, SLH-DSA-SHAKE-256f | 32 | 64 | 128 | 49,856 | 5 |
Bytes. Source: NIST FIPS 205, Section 11, Table 2; private key length from Section 3.1 (4n bytes).
Falcon (planned FN-DSA, FIPS 206) sizes
FN-DSA will be NIST's standard derived from Falcon. As of NIST IR 8547 (October 2026), NIST says it "is developing" that FIPS, so these sizes come from the Falcon team's own specification (version 1.2) and may change in the final standard. Source: Falcon specification, Section 3.13, Table 3.3 (public key, signature, target level).
| Parameter set | Public key | Private key | Signature | Target NIST level |
|---|---|---|---|---|
| Falcon-512 | 897 | 1,281 | 666 | I (1) |
| Falcon-1024 | 1,793 | 2,305 | 1,280 | V (5) |
Bytes. Source: Falcon specification v1.2, Table 3.3. Signature is the padded length (sbytelen). Private key sizes are calculated from the encoding rules in Section 3.11.5 (header byte, then f and g at 6 bits per coefficient for degree 512 or 5 bits for degree 1024, then F at 8 bits). The Falcon site describes the private key as about three times the signature.
RSA and ECC sizes for comparison
These are the quantum-vulnerable algorithms being replaced. NIST IR 8547 (Section 4.1) deprecates all of them after 2030 and disallows them after 2035.
| Algorithm | Public key | Private key | Signature or shared secret | Source |
|---|---|---|---|---|
| RSA-2048 | 256 (modulus) plus exponent | Varies by encoding | 256 (signature) | RFC 8017, Section 8.1.1: signature length equals modulus length |
| RSA-3072 | 384 (modulus) plus exponent | Varies by encoding | 384 (signature) | RFC 8017, Section 8.1.1 |
| ECDSA P-256 | 65 (uncompressed point) | 32 (256-bit scalar) | 64 (signature, r and s) | RFC 8446, Section 4.2.8.2 (point); RFC 7518, Section 3.4 (ES256 signature) |
| Ed25519 | 32 | 32 | 64 (signature) | RFC 8032, Section 7 |
| X25519 | 32 | 32 | 32 (shared secret) | RFC 7748, Sections 5 and 6.1 |
Bytes. The P-256 private key size is our calculation: a scalar below a 256-bit group order fits in 32 bytes. RSA private keys hold several values and their size depends on the encoding format.
Side by side: signatures and key exchange
Signature plus public key is what a verifier receives in a certificate or token. Totals below are our arithmetic from the tables above.
| Signature scheme | Public key + signature | vs Ed25519 (96) |
|---|---|---|
| Ed25519 | 96 | 1x |
| ECDSA P-256 | 129 | 1.3x |
| RSA-2048 | 512 (modulus + signature) | 5.3x |
| Falcon-512 | 1,563 | 16x |
| ML-DSA-44 | 3,732 | 39x |
| ML-DSA-65 | 5,261 | 55x |
| ML-DSA-87 | 7,219 | 75x |
| SLH-DSA-128s | 7,888 | 82x |
| SLH-DSA-256f | 49,920 | 520x |
For key exchange, TLS sends the hybrid of a classical and a post-quantum share. RFC 10024 (August 2026), Sections 4.1 and 4.2, gives the exact sizes:
| TLS 1.3 group | Client key share | Server key share |
|---|---|---|
| X25519 (classical) | 32 | 32 |
| X25519MLKEM768 | 1,216 | 1,120 |
| SecP256r1MLKEM768 | 1,249 | 1,153 |
| SecP384r1MLKEM1024 | 1,665 | 1,665 |
Bytes. Source: RFC 10024, Sections 4.1 and 4.2; X25519 from RFC 8446, Section 4.2.8.2.
What the sizes mean in practice
TLS handshakes
Key exchange is the cheap part. Chrome measured a 4% slowdown in TLS handshake time from the extra 1.1 kB server-to-client and 1.2 kB client-to-server that hybrid key agreement adds, as Cloudflare reports. That cost has been accepted. Chrome 131 switched to X25519MLKEM768 (Google). Apple's OS 26 releases advertise it on TLS connections automatically (Apple), and over two-thirds of browser traffic to Cloudflare was post-quantum encrypted in June 2026 (Cloudflare).
Certificate chains
Signatures are the expensive part. Cloudflare counts at least five signatures and two public keys in a typical TLS handshake, not counting an OCSP staple. Replacing them with ML-DSA-44 adds 14.7 kB (two 1,312-byte keys plus five 2,420-byte signatures). Cloudflare's 2021 experiment saw client and middlebox failures rise steeply above 10 kB of added certificate data, and a handshake slowdown of about 15% below 9 kB. With ML-DSA-65 the same seven objects total 20,449 bytes, by our arithmetic from FIPS 204.
That is why Cloudflare and Chrome are testing Merkle Tree Certificates, which replace most of those signatures with a Merkle tree proof under 800 bytes in the common case (Cloudflare, October 2025). Falcon-512, at 1,563 bytes for key plus signature, is the other way out once FN-DSA is final.
WebAuthn and passkeys
A passkey assertion today carries a 64-byte ES256 or Ed25519 signature. RFC 9964 (May 2026) registers ML-DSA-44, ML-DSA-65 and ML-DSA-87 as COSE algorithms -48, -49 and -50, the identifiers WebAuthn uses to name credential algorithms. An ML-DSA-44 assertion would carry 2,420 bytes of signature, and the credential public key stored by the relying party grows from 32 or 65 bytes to 1,312. Authenticator storage and the size of the registration response grow with it.
Blockchain transactions
Bitcoin's Taproot uses 64-byte Schnorr signatures and 32-byte public keys (BIP-340). An ML-DSA-44 signature and key together are 3,732 bytes, about 39 times the 96 bytes a key path spend relies on, and every byte of witness data costs fees. The current quantum proposal, BIP-360 (status: Draft), removes the exposed key path instead of adding post-quantum signatures, and its authors say post-quantum signatures would need a separate future proposal.
NIST security categories
NIST rates post-quantum schemes against the cost of breaking a reference symmetric primitive rather than in "bits of security". Source: NIST IR 8547, Section 4.1, Table 1.
| Category | At least as hard to break as | Example | Parameter sets claimed |
|---|---|---|---|
| 1 | Key search on a 128-bit block cipher | AES-128 | ML-KEM-512, SLH-DSA-*-128s/f, Falcon-512 (target) |
| 2 | Collision search on a 256-bit hash | SHA-256 | ML-DSA-44 |
| 3 | Key search on a 192-bit block cipher | AES-192 | ML-KEM-768, ML-DSA-65, SLH-DSA-*-192s/f |
| 4 | Collision search on a 384-bit hash | SHA3-384 | None of the FIPS 203 to 205 sets |
| 5 | Key search on a 256-bit block cipher | AES-256 | ML-KEM-1024, ML-DSA-87, SLH-DSA-*-256s/f, Falcon-1024 (target) |
Sources: NIST IR 8547, Table 1; claims from FIPS 203 (Section 8), FIPS 204 (Table 1), FIPS 205 (Table 2) and the Falcon specification (Table 3.3). CNSA 2.0 for US National Security Systems requires category 5 sets, ML-KEM-1024 and ML-DSA-87, per an NSA presentation at RSAC 2026.
Status and dates
| Algorithm | Standard | Status (November 2026) | Source |
|---|---|---|---|
| ML-KEM | FIPS 203 | Final; published and effective August 13, 2024 | FIPS 203 |
| ML-DSA | FIPS 204 | Final; published and effective August 13, 2024 | FIPS 204 |
| SLH-DSA | FIPS 205 | Final; published and effective August 13, 2024 | FIPS 205 |
| FN-DSA (Falcon) | FIPS 206 (planned) | In development at NIST; not yet approved | NIST IR 8547, Section 2.1.1 |
| HQC | FIPS not yet numbered | Selected March 11, 2025; FIPS under development | NIST, NIST IR 8547 |
| RSA, ECDSA, EdDSA, DH, ECDH | FIPS 186, SP 800-56A/B | Deprecated after 2030, disallowed after 2035 | NIST IR 8547, Section 4.1 |
| Hybrid TLS (X25519MLKEM768) | RFC 10024 | Standards Track RFC, August 2026 | RFC 10024 |
| ML-DSA in X.509 | RFC 9881 | Standards Track RFC, October 2025 | RFC 9881 |
| ML-DSA in JOSE and COSE | RFC 9964 | Standards Track RFC, May 2026 | RFC 9964 |
US federal deadlines sit inside NIST's window. Per NIST IR 8547, Executive Order 14412 and OMB Memorandum M-26-15 require high value assets and high-impact systems to adopt post-quantum key establishment by December 31, 2030 and post-quantum signatures by December 31, 2031. The migration guide covers the full timeline, including CNSA 2.0.
Frequently asked questions
How big is an ML-DSA signature?
2,420 bytes for ML-DSA-44, 3,309 bytes for ML-DSA-65 and 4,627 bytes for ML-DSA-87, per FIPS 204, Table 2.
How big is an ML-KEM-768 public key?
The encapsulation key is 1,184 bytes and the ciphertext is 1,088 bytes, per FIPS 203, Table 3. In hybrid TLS the client share is 1,216 bytes with X25519 added.
Which post-quantum signature is smallest?
Falcon-512, at a 666-byte signature and 897-byte public key, but FN-DSA is not yet a final NIST standard. Among final standards, ML-DSA-44 has the smallest signature (2,420 bytes) and SLH-DSA-128s the smallest public key (32 bytes).
Why are SLH-DSA signatures so large?
SLH-DSA builds signatures from many hash-based one-time and few-time signatures in a tree. The trade is conservative security, resting only on hash functions, for signatures of 7,856 to 49,856 bytes.
Related reading
- Post-quantum cryptography for authentication: the enterprise migration guide
- The future of hashing: quantum resistance and beyond
- Choosing the right hash algorithm: a decision framework
- Top post-quantum cryptography tools
Deepak Gupta is the Co-founder and CEO of GrackerAI and an AI and Cybersecurity expert with 15+ years in digital identity and enterprise security. He founded a CIAM platform and scaled it to serve over one billion users. He writes about cybersecurity, AI, and B2B SaaS at guptadeepak.com.
More from Deepak Gupta
Every page on guptadeepak.com is hand-curated by Deepak Gupta. Pick a thread:
- About Deepak Gupta Founder, cybersecurity architect, and writer at guptadeepak.com.
- My journey From LoginRadius (2013, 1B+ users) to GrackerAI, in milestones.
- Publications & patents Books, free e-books, a journal special issue, and five granted patents.
- Research Hub Curated research, buyer's guides, vendor comparisons, and technical deep-dives.
Get the newsletter
New writing on identity, AI security, and building software, delivered when it ships. No tracking pixels, no funnels, unsubscribe with one click.