What’s new at IBM Quantum: Q3 2026
Error-correction tools, Nighthawk r2 access and the latest Qiskit updates.

A front-row seat to the machines, discoveries,
and decisions bringing quantum closer.
Helios · error-corrected demonstration
Helix · relative to physical Clifford error
IBM Nighthawk r2 · reported benchmark
Production wallet-key recovery
Follow 31 reports across hardware, research, and crypto security. Search, filter, and build your reading list.
2025 was an announced goal; 2027–28 are roadmap targets. The 2026 benchmark uses error mitigation. This is not a measured annual growth rate.
Announcements, research, and the developments worth following.
Error-correction tools, Nighthawk r2 access and the latest Qiskit updates.
Stage C brings independent validation of proposed systems. It is not a demonstrated utility-scale result.
Selected research and commercial users can test error-aware programming based on D-Wave's dual-rail gate-model architecture. The beta provides simulator access while the corresponding gate-model hardware remains forthcoming.
Quantinuum reports a 4.28× logical improvement over physical Clifford error, without post-selection.
The center provides Majorana 2 hardware for independent DARPA testing.
Selected systems and research platforms. Every number needs context.
Physical qubits, logical qubits and trapped-atom arrays measure different things. These cards are not a leaderboard.
A selected, source-linked record of meaningful progress.
Quantinuum reports a logical compute error rate below its physical Clifford baseline without postselection. Its separate 64-qubit result uses error detection; it is not the same measurement as 48 error-corrected logical qubits.
2.8 × 10⁻⁴ logical compute errorIBM reports accurate observable estimation on circuits exceeding 7,500 gates and throughput above 100,000 circuits per second. The processor has 120 programmable qubits; its 458 total quantum elements also include couplers and reset elements.
7,500+ gates · 100,000+ circuits/sA Nature study demonstrates reinforcement learning that uses error-detection events to adjust Willow control parameters during quantum memory operation. This addresses calibration drift; scaling studies on larger systems remain simulations.
Adaptive control during error correctionThe Foundation reports initial Falcon implementations from Anza and Firedancer and describes a future wallet migration path. This documents preparation and research, not universal post-quantum protection already active for Solana accounts.
Falcon prototypes · migration planningTheoretical circuits could solve 256-bit elliptic-curve discrete logarithms with fewer than 500,000 physical superconducting qubits under specified error and connectivity assumptions. This is not demonstrated hardware. April 15 revision patches proof-system software.
<500,000 physical qubits, modeledThe cryptographic risk, the evidence, and the race to migrate. Solana first.
For an ordinary keypair wallet, knowing its Solana address reveals the public key. A future capable quantum attacker would not need to wait for its first outgoing transaction.
The tracked results do not demonstrate recovery of a production wallet key. A hardware milestone or a theoretical attack estimate is not a completed attack.
PDAs have no private key. Their security depends on the controlling program, its authorization rules and any upgrade or recovery keys. A PDA is not automatically a post-quantum wallet. Source ↗
In April 2026, the Solana Foundation reported initial Falcon implementations from Anza and Firedancer and a plan to migrate wallets if needed. This is preparation, not network-wide post-quantum activation.
Google’s 2026 model for secp256k1, assuming superconducting hardware, 10⁻³ physical error rates and planar connectivity. Modeled runtime: minutes.
This is one modeled circuit option, not an available machine. It is not a measured Ed25519/Solana threshold.
Read the resource estimatePublished attack models and prototype defenses. No attack date follows from these results.
An internal planning deadline, not a prediction that crypto breaks in 2029.
Arrival date unknown. Requires a capable machine, an exposed key, and authorization that has not migrated.
Offline key storage protects against many conventional threats. It does not change Ed25519 or secp256k1: an attacker who can derive signing power from the public key would not need the physical device. This is an inference from the public-key threat model above.
A quantum-resistant launch signature can protect proof of origin. Protecting holdings requires the spending authorization itself to enforce post-quantum checks, with no vulnerable alternate authorization path. A recorded signature fingerprint alone does not establish that protection; any specific implementation needs verification.
That outcome is not inevitable. The relevant race is between cryptographically capable hardware and migration of the network, wallets and applications. Solana has published preparation work. A future key-recovery capability could threaten unmigrated authorities without automatically destroying every account, program or asset.
Solana’s published preparation work ↗Educational threat models, not a wallet audit. No wallet connection is needed. Exposed public keys are not evidence of present-day theft; reviewed metrics are not a price forecast.
Quantum Watcher is an independent, curated observatory. It is not affiliated with the companies tracked here.
Every milestone links to a research paper or the organization’s own report. Provider claims are labeled.
Hardware scale, logical performance, projections and resource estimates stay separate.
Nine automatic source connections refresh on demand with a 15-minute cache. The Wire shows source health; reviewed metrics remain dated 8 Oct 2026.
Coverage is selective, not exhaustive. News does not automatically change verified metrics. Saved links stay in this browser.