Quantum Error Correction Tests Begin on Israel's Arbel Processor

What happened

Quantum X Labs, an Israeli quantum technology company, has started what it calls an end-to-end quantum error correction (QEC) program on Arbel, a superconducting quantum processor at the Israeli Quantum Computing Center (IQCC). Quantum Machines operates the center.11 The company announced the program on October 26, 2026. Hours later Gix Internet, which owns about 13% of Quantum X Labs (2,818,585 shares), put out its own release about it.122

The first hardware experiments are meant to lay the groundwork for a single workflow with three parts. The first part characterizes and calibrates the processor to find its particular noise patterns. The second feeds that data into how quantum circuits are controlled and run. The third uses a "noise-aware" decoder to read error signals and work out what correction to apply.12 The next phase will add that decoder directly to the working pipeline. The longer-term goal is a closed loop, in which error information from the processor becomes corrective action with as little delay as possible.11

The main point gets lost in most write-ups. The company itself says this first stage does not show real-time error correction or fault-tolerant computation, and that both are goals for later phases.1113 Despite the word "milestone" in the release, this is the start of an experimental program, not a breakthrough.

Reading the coverage

The reporting matches closely on the facts, mostly because nearly all of it comes from the same press release. Financial outlets reran the release with little change.1215 Trade sites summarized it and kept the company's description of a "key first hardware milestone."1720 Coverage differs mainly in tone. Automated summaries on investor platforms put the disclaimer up front. One headlined its write-up with the fact that real-time correction is still a later goal. Another called the company's market-moving impact "moderate" and its sentiment neutral.13 A second automated summary listed the lack of a real-time demonstration as its main negative and said it showed the technology is "still in a nascent stage."18

A Yahoo Tech report gave the most measured account. It said the first experiments are meant to inform later stages, not to show a finished system, and that the company has given no date for a working closed-loop setup.14 One trade write-up went the other way and called the hardware-first approach a departure from efforts that lean on simulation.17 That claim needs context. Testing decoders on real hardware data is common across the field, as shown by Quantum X Labs' own earlier benchmarks on Google's public surface-code dataset.1

There is also a small discrepancy about the company's listing. Yahoo identified it by an over-the-counter ticker, QUTX.14 The company's own release and Gix's corporate site use the Nasdaq symbol QXL, and Gix says the shares have traded on Nasdaq since April 30, 2026.118 This looks like a stale identifier and doesn't change the substance of the news. Still, it shows how thinly this small-cap company has been covered.

The decoder story behind the hardware story

The Arbel announcement follows an October 22 update from Gix about two earlier Quantum X Labs results.1 The company said its QEC decoder now runs on NVIDIA's CUDA-Q platform. On 10,000 held-out runs from Google's public surface-code data, it made 12.7% fewer logical errors than the widely used PyMatching decoder at code distance 3, and 12.1% fewer at distance 5.1 In a separate 25-round streaming test, where error data arrived one round at a time, its logical error rate was about 6% lower at distance 3 and 9% lower at distance 5.1 The company said its next step is to run the decoder in a real-time pipeline over an NVIDIA NVQLink connection.1

Gix's security unit, QuantumQ Security, also reported results for a calibration technique called LayerQake on three IBM processors. It said the technique cut measurement bias from roughly 11–15% to under 6% and reduced maximum cumulative error by 15 to 21 times.19 Gix said plainly that it had not independently checked these technical results and that it does not control Quantum X Labs.16

Together, the two announcements show a clear strategy. Quantum X Labs is not building qubits. It is building the classical software that sits next to them, which reads error signals quickly and accurately enough to keep a computation running. The company describes this as technology meant to work across different hardware types, combining software, algorithms and accelerated classical computing.11 Arbel is where the decoder, so far tested only on recorded data, is supposed to start working with a live processor.

Why decoding is a real bottleneck

The focus on latency is justified. Fault-tolerant machines need error information to be processed fast enough to keep up with the computation, a point the Arbel release makes too.12 NVIDIA's NVQLink was built for this kind of tight coupling. It is listed as offering up to 400 Gb/s of GPU-to-quantum-processor throughput and an FPGA-GPU-FPGA round trip of under 4 microseconds.26 Specialists such as Riverlane have built businesses around real-time decoding hardware, which shows the classical side of QEC has become a competitive area in its own right.24

A 12% improvement over PyMatching on a public benchmark is useful and believable. It is not transformative. In this field, decoder accuracy matters only if it holds up under the timing limits of a live machine. That is why the Arbel program is the more important test, and why its lack of results so far is the key detail.

Where the industry's qubit roadmaps stand

The wider 2026 picture puts the announcement in perspective. Quantinuum said on October 19 that its Helios system reached 99.9995% logical fidelity. It called that ten times better than its own September record and 100 times better than the best published result on any other type of hardware.22 A Sandia National Laboratories researcher who was not involved said the logical qubit was nearly 100 times better than its physical parts.22 QuEra reported up to 96 verified logical qubits from 448 neutral atoms in a January 2026 Nature paper, at about 4.7 physical qubits per logical qubit.21

Even the leaderboards disagree on definitions. One tracker credits Quantinuum's Helios with 50 error-corrected logical qubits.26 Another lists 48 as error-detected.24 A third gives 94 error-detected and 48 fully error-corrected.21 Infleqtion's 30 logical qubits, announced in September, met its 2026 target. One analyst noted that they run in a distance-2 code with post-selection, which means errors are detected but not corrected. The same analyst pointed out that the company's original 2026 goal had specified error correction.23 Infleqtion now targets 100 logical qubits in 2028 and 1,000 in 2030. IBM's roadmap aims for its fault-tolerant Starling system, with 200 logical qubits, in 2029.26

The common thread is that the line between detecting and correcting errors, and between demonstrating and delivering, has become where credibility is won or lost. Quantum X Labs gets credit here for stating what its Arbel work does not yet show.

The verdict

This should be read as a capability-building step by a small company trying to establish itself in the QEC software layer, not as a quantum computing breakthrough. Its strengths are real access to a superconducting processor through an Israeli national facility, a decoder with reasonable benchmark results, and a stated plan to work with control-system and accelerated-computing partners.111 Its weaknesses are just as clear. There are no hardware results yet, no timeline, and the earlier technical claims have not been independently verified.141

The fact that a 13% shareholder announced the news on top of the company's own release suggests both firms want investors to see it. That is expected for small listed companies, but it calls for caution.219 What will matter is the next phase, when the noise-aware decoder runs inside a live loop on Arbel and the company publishes latency and logical error figures. Until then, this is a program launch and nothing more.