As of 2026-07-10Level 4

qLDPC Codes

Quantum low-density parity-check (qLDPC) codes protect many logical qubits with far fewer physical qubits than surface codes: IBM's bivariate-bicycle 'gross code' [[144,12,12]] stores 12 logical qubits in 288 physical qubits — matching surface-code performance with about 10x fewer qubits.

What it means

The surface code encodes just one logical qubit per patch, so a useful machine would need millions of physical qubits.qLDPC codes attack this overhead.'Low-density' means each parity check involves only a few qubits and each qubit participates in only a few checks — but unlike the surface code, the checks may connect distant qubits, and that long-range structure buys a much higher encoding rate.The landmark result: IBM's bivariate-bicycle 'gross code' [[144,12,12]] (Bravyi et al., Nature 627, March 2024) encodes 12 logical qubits at distance 12 using 144 data plus 144 check qubits = 288 physical qubits, matching surface-code performance with roughly 10x fewer qubits.The trade-offs are real: the required long-range couplers are hard to build in superconducting hardware, decoding is heavier, and logical operations are less mature than surface-code lattice surgery.IBM's roadmap builds on the gross code: Kookaburra (2026, first qLDPC memory module), Starling (2029, 200 logical qubits and 100 million gates), Blue Jay (2033, 2,000 logical qubits).These roadmap entries are plans, not achievements — dates can slip and targets can change.Meanwhile the classical side is catching up: the first real-time qLDPC decoding was demonstrated at about 67 µs median latency using an NVIDIA GH200 attached to Quantinuum's Helios (NVIDIA Technical Blog, 2025-11-17).

Everyday analogy

Guarding 12 treasures at school: you could give every treasure its own huge team of 100 guards (surface code) — or set up one clever neighborhood watch where each guard checks only a few lockers and each locker is checked by a few guards, some patrolling across the whole school. That shared watch protects all 12 treasures with far fewer guards.
qLDPC is the quantum cousin of classical LDPC codes — the error-correcting codes already inside your phone's Wi-Fi and 5G. The classical version took decades from invention to deployment; the quantum version is in the middle of that journey.

Common misconceptions

  • '10x fewer qubits' is not free: the gross code needs long-range connections between distant qubits (hard in 2D superconducting chips, where the surface code only needs nearest neighbors), heavier real-time decoding, and its logical-gate toolbox is less mature than surface-code lattice surgery.
  • IBM's Kookaburra (2026), Starling (2029) and Blue Jay (2033) are roadmap PLANS, not demonstrated machines — treat the dates as intentions, not facts.
  • qLDPC does not replace the threshold theorem or make errors disappear — it reduces the physical-qubit overhead per logical qubit; error correction is still continuous, and decoding must keep up in real time.

Key takeaways

  • qLDPC codes use sparse parity checks with long-range connections to encode many logical qubits per block, unlike the one-per-patch surface code.
  • IBM's bivariate-bicycle 'gross code' [[144,12,12]] (Bravyi et al., Nature 627, March 2024): 12 logical qubits at distance 12 in 144 data + 144 check = 288 physical qubits — about 10x fewer than comparable surface codes.
  • The price: long-range couplers, heavier decoding, and less mature logical operations than surface-code lattice surgery.
  • IBM's roadmap (Kookaburra 2026 → Starling 2029, 200 logical qubits / 100M gates → Blue Jay 2033, 2,000 logical qubits) is a plan, not an achievement.
  • First real-time qLDPC decoding: ~67 µs median, NVIDIA GH200 + Quantinuum Helios (NVIDIA Technical Blog, 2025-11-17).

Check your understanding

What does the code notation [[144,12,12]] of IBM's gross code mean?

  1. A.144 logical qubits, 12 physical qubits, 12 gates
  2. B.144 data qubits encode 12 logical qubits with code distance 12
  3. C.144 checks run 12 times over 12 qubits
  4. D.A code that corrects exactly 144 errors
Show the answer

Answer: B. 144 data qubits encode 12 logical qubits with code distance 12

Why: In [[n,k,d]] notation, n=144 data qubits encode k=12 logical qubits at distance d=12. With 144 additional check qubits, the gross code uses 288 physical qubits — about 10x fewer than surface codes of comparable performance.

What is the main hardware price of the gross code's ~10x qubit savings compared with the surface code?

  1. A.It only works at room temperature
  2. B.It requires long-range connections between distant qubits, while the surface code needs only nearest-neighbor links
  3. C.It cannot detect phase errors
  4. D.It needs no classical decoding
Show the answer

Answer: B. It requires long-range connections between distant qubits, while the surface code needs only nearest-neighbor links

Why: The gross code's high encoding rate comes from checks that connect distant qubits — hard to build in 2D superconducting chips. It also demands heavier real-time decoding, and its logical operations are less mature than surface-code lattice surgery.

Builds on

Quantum Error CorrectionLogical Qubit

Graded 2026-07-10 (human sign-off): timebound, asOfDate 2026-07-10 — IBM roadmap dates (Kookaburra 2026, Starling 2029, Blue Jay 2033) are corporate plans, not achievements, and may slip; the ~67 µs real-time decoding record (NVIDIA GH200 + Quantinuum Helios, Nov 2025) is a fast-moving benchmark. The [[144,12,12]] gross-code result itself is peer-reviewed (Nature 627, 2024).

Learn it hands-on

This concept is part of a 46-level curriculum with an interactive simulator and Lumen, a tutor whose answers are verified before you see them. Levels 1–5 are free.

Start freeSee the curriculum