Error correction / Z-QEC
Platinum Z-QEC
Quantum error correction tuned to your chip, for hardware makers
Z-QEC is our own decoder. Platinum and Gold are both tuned to your chip, Platinum to our highest accuracy. Both are agreed under contract and measured on your chip's own held-out shots, so the gain you see is one measured on your hardware.
Measured on public Willow chip data
13 to 27%
fewer logical errors per cycle than Sparse Blossom matching, Z-QEC Gold with no chip data like it, d3 to d7, the 2026 Willow chip
5 to 10%
fewer than Tesseract, calibrated, which runs on the same tuned noise file, d3 to d7, the 2026 Willow chip
At least 15%
fewer errors than Libra at d7, on held-out shots of the 2024 Willow chip
Level
with AlphaQubit 2 on Lambda, the 2026 Willow chip
For a maker, fewer logical errors per cycle means a lower error at the same code distance, or the same error from fewer physical qubits. Your own gain is measured on your chip's held-out shots.
Tuned to your chip, highest accuracy
Z-QEC Platinum
- Z-QEC tuned to your chip, at our highest accuracy
- Surface and repetition code memory on your hardware
- Measured on your own held-out shots, against your current decoder on the same shots
- No promised figure before your data is measured
- Real time decoding on demand
Tuned to your chip
Z-QEC Gold
- Z-QEC tuned to your chip's own data
- Surface and repetition code memory on your hardware
- Measured on your own held-out shots, against your current decoder on the same shots
- No promised figure before your data is measured
- Real time decoding on demand
Which one suits you
| Z-QEC Platinum | Z-QEC Gold | Console error correction | |
|---|---|---|---|
| For | Hardware makers, under contract | Hardware makers, under contract | Anyone, self-serve |
| Decoder | Z-QEC | Z-QEC | PyMatching |
| Tuned to your chip | Included | Included | Not included |
| Our highest accuracy | Included | Not included | Not included |
| Runs on your chip's own data | Included | Included | Not included |
| Measured on your held-out shots, against your current decoder | Included | Included | Not included |
| Logical error rate with a 95% interval | Included | Included | Included |
| ZQEC-v0.1 certificate | Included | Included | Included |
| Real time decoding on demand | Included | Included | Not included |
| Price | Contact us | Contact us | Per job, in the console |
What you receive
- The logical error per cycle at each code distance, and Lambda, on your held-out shots with 95% intervals
- The same figures for your current decoder on the same shots, so the gain is measured on your chip
- A ZQEC-v0.1 certificate for the logical error rate
- Real time decoding on demand
Where it stands today
On public data from the 2024 Willow chip, Z-QEC Platinum has the highest Lambda of any decoder, 2.21 on held-out TEST shots, and makes at least 15% fewer errors than Libra at d7 on the same shots. Against the Willow neural network, trained on the chip's own data, Platinum is level at d7. On the 2026 Willow chip, Platinum and Gold are level with AlphaQubit 2 on Lambda, and Platinum is near level at d3. Z-QEC Gold with no chip data has the lowest d3 error of any decoder with no chip data.
All figures come from offline decoding.
Lambda on the 2024 Willow chip
- Z-QEC Platinum, TESTZKSF, scored by us2.18 to 2.25
- Z-QEC Gold, TESTZKSF, scored by us2.15
- Willow neural networkAs published2.14
- LibraAs published2.04
- Correlated matchingRiverlane, as published2.02
Lambda on the 2026 Willow chip
- AlphaQubit 2Scored by us2.624
- Z-QEC Gold, VALZKSF, scored by us2.59 to 2.65
- Z-QEC Platinum, VALZKSF, scored by us2.56 to 2.62
- Tesseract, calibratedScored by us2.509
- Sparse Blossom, SI1000Scored by us2.196
Higher is better. A white line marks a range or a 95% interval
Other chips
On IBM's Nighthawk surface code, Z-QEC Gold made 7% fewer logical errors than matching on the same shots, level with BeliefMatching. On five more chips running a repetition code, it was level with or ahead of matching on every one, and never wrong on a shot that matching decoded right.
| Chip | Test | Shots | Z-QEC Gold | Matching |
|---|---|---|---|---|
| Nighthawk ibm_miamiSurface code, d5, 2 to 8 rounds | Surface code, d5, 2 to 8 rounds | 12,288 | 1,946 | 2,090 |
| TianyanRepetition code, d3 to d11 | Repetition code, d3 to d11 | 102,400 | 38 | 46 |
| BrisbaneRepetition code, d3 to d11 | Repetition code, d3 to d11 | 102,400 | 27 | 31 |
| WuyueRepetition code, d3 to d9 | Repetition code, d3 to d9 | 81,920 | 102 | 105 |
| TorinoRepetition code, d3 to d11 | Repetition code, d3 to d11 | 102,400 | 9 | 10 |
| SherbrookeRepetition code, d3 to d11 | Repetition code, d3 to d11 | 51,200 | 6 | 6 |
Logical errors on the same shots, lower is better, with each decoder given the same published chip data. BeliefMatching made about 1,993 on Nighthawk, counted from the dataset's own figures. A brief test with small counts, on public data alone; the repetition code ran one round, on validation shots.
Data: Nighthawk ibm_miami from Drouet et al., Zenodo record 21336494; Tianyan, Wuyue and IBM's Brisbane, Torino and Sherbrooke from He Songhuan's PAEMS dataset, Zenodo record 18230210. Both under CC BY 4.0. Changes: scored by us. Names identify the data and imply no partnership or endorsement
Talk to us
Tell us about your chip and the codes you run. TEST data for the published figures is available on request. How Z-QEC works stays private, and anything beyond the figures is discussed under a non-disclosure agreement