HostNexFlow › Library › Ledger vs Tangem
Ledger vs Tangem: the screened wallet with a seed, and the card with neither
Every hardware wallet answers two questions — where do keys live, and how do you verify what you sign. Ledger answers with a secure element plus a screen plus a seed phrase. Tangem answers with an EAL6+ card, firmware that can never update, and — by default — no seed at all.
This is the most philosophically opposed pair in hardware custody. Ledger built the screened, updatable, seed-anchored device that became the industry default — and absorbed the industry's loudest criticism along the way. Tangem bet the opposite on every axis: no screen, no seed, no updates, no company-held recovery — a card you tap and a key that physically cannot leave the chip.
Every row below names its source and date.
The parameter table
| Parameter | Ledger | Tangem |
|---|---|---|
| Form factor | Nano S Plus / X / Flex / Stax — small devices with a screen and buttons; USB-C / Bluetooth. | A bank-card-shaped smartcard (or ring) — no screen, no battery, no cable; NFC tap to your phone is the whole interface. |
| Key model | Seed phrase by default — BIP-39, portable to any wallet; Recover option adds a sharded cloud backup. | Seedless by default — the key is generated on-card by a hardware RNG and cloned to backup cards; a seed-phrase mode exists but the identity is seed-free. |
| Verification | Trusted display: what you sign is shown on the device screen — the point of a screen. | No screen — the phone app displays the transaction; you trust the app's rendering, which is the honest criticism the design accepts. |
| Firmware | BOLOS on a secure element — updatable, which enables both improvements and (per critics) the Recover-style attack surface. | Burned once at manufacture — never updatable, by design: no malicious-update path exists, and no security patch path exists either. |
| Audits & openness | Closed firmware on the SE; Donjon red-team lab; apps open. The industry's largest install base. | Firmware closed (“security through obscurity” — Tangem's own words), audited twice: Kudelski 2018 and Riscure 2023; the companion app is open-source. |
| Company & record | French, founded 2014; Jul-2020 commerce breach (~272K records → phishing); Dec-2023 ConnectKit supply-chain incident (~$600K, <2h); no wallet-firmware breach of keys. | Swiss (Zug), founded 2017; claims 6M+ devices shipped without a compromised card; no customer-data megabreach on file — a thinner corporate attack surface partly because it holds less data. |
| Loss model | Lose the device → recover by seed (or Recover service). Physical card loss is secondary. | Lose one card → the other cards in the set still hold the key; lose all of them without a seed phrase → funds are gone. Redundancy is physical multiplication. |
The seed-versus-seedless fork
Ledger's model inherits BIP-39: one mnemonic generates everything, recoverable forever on any compatible wallet, and losable forever by one photograph or one house fire. The May-2023 Recover controversy showed where that path leads — an optional paid service that shards an encrypted seed to third-party custodians. Optional, audited, and still the angriest week in Ledger's history, because it proved the firmware could ship keys out if asked.
There is also a custody-theory dimension the marketing skips: a BIP-39 seed is an open standard — the escape hatch is that your backup works anywhere, on any vendor's device, forever. Tangem's card-clone model is proprietary redundancy — your backup only works because another Tangem card holds it, which quietly makes Tangem a permanent dependency of your own recovery path. If the company vanished tomorrow your cards still sign, but your ability to re-create the backup chain dies with the ecosystem.
Tangem removes the question: in the default flow no seed exists — the on-chip RNG generates the key inside the EAL6+ element, and “backup” means cryptographically cloning the key to the other cards in your set. Nothing is ever written down, so nothing can be photographed, phished, or typed into a fake site. The cost is symmetric: no seed means no escape hatch — lose every card and the funds are unrecoverable by anyone, including Tangem.
The screen question is not cosmetic
A hardware wallet's screen is a trust boundary: it's the one place a transaction can be shown that malware on your computer or phone cannot fake. Ledger devices display the destination and amount on the secure element's own display before you sign — blind-signing on the host is the failure the screen exists to prevent.
Tangem accepts that you are signing what the phone app shows. The NFC channel carries the unsigned transaction from a potentially compromised phone to the card — the card verifies nothing about what it signs beyond the access code. Tangem's honest answer is that screenless signing is safe enough for a phone-first wallet, and the audits (Kudelski 2018, Riscure 2023) cover the firmware's integrity, not the host's honesty. That's a real trade-off, and heavy users should understand it before choosing the card.
Updatable vs immutable firmware
Ledger's BOLOS updates — new coin support, security patches, Recover. An updatable device trusts the vendor forever: every future signed firmware is a potential key-exfiltration path in the eyes of the threat-model purists, and the 2023 Recover debate is that argument made real.
Tangem burned its firmware into the chip once and made it non-updatable: the vendor literally cannot push a malicious update — the strongest possible vendor-trust answer, priced against the impossibility of fixing a flaw after ship. It also means the audits carry the whole assurance burden, since nothing can be patched later. Two opposite answers to “what if the vendor turns evil” — Ledger says trust us, with a screen; Tangem says nobody can update anything, ever.
There is a third axis nobody advertises: supply-chain trust. A Ledger ships as electronics you inspect; a Tangem ships as a card you cannot inspect, and its app verifies firmware authenticity against burned-in fingerprints — you are trusting Samsung's chip fabrication and Tangem's provisioning line end-to-end. Both companies publish enough to audit the claim; neither is fully trustless, which is the honest constant under every hardware wallet ever sold.
Who should pick which
Pick Ledger if you want the screened signing check, BIP-39 portability, the broadest asset and app support, and you can tolerate a vendor capable of pushing firmware. Pick Tangem if seed phrases are your actual failure mode — no phrase to lose or photograph — and you're comfortable trading the trusted screen for NFC tap signing and immutable firmware. The sophisticated answer many arrive at: Tangem for the phone-pocket daily wallet, a screened device for the vault-sized signing decisions.
Frequently asked
Is Tangem safe without a screen?
The key custody is strong (EAL6+, two audits); the trade is that you sign what the phone app shows — a compromised phone could display one transaction and send another. Screened devices (Ledger, Trezor) exist precisely to close that gap.
Does Tangem really never update?
Correct — firmware is burned at manufacture and cannot update, deliberately: no malicious-update path exists. The cost is no patch path either; the audits carry the whole assurance burden.
What happens if I lose my Tangem card?
Nothing, if you backed up to the other cards in the set — the key is cloned across them. Lose every card without a seed phrase and the funds are unrecoverable by anyone, including Tangem.
Can I use a seed phrase with Tangem?
Yes — second-generation wallets offer a seed-phrase mode (BIP-39 import/generation). The default and identity-defining mode is seedless.
Is Ledger's Recover a risk?
It's optional and off by default — but its 2023 existence proved the firmware can export seed shares when asked, which is why the purist critique focuses on updatable firmware rather than the service itself.
Which survived incidents better?
Different incident classes: Ledger's wounds are corporate/supply-chain (2020 commerce breach, 2023 ConnectKit) — no key-level breach. Tangem has no comparable breach on file; its 6M+ device record is clean but younger and smaller-scale.