The Signal
Moth claims Tessa record for quantum data encoding and readback
Moth says its Tessa method encoded and recovered 1,568 values on an IBM quantum processor. A preprint reports that recovery quality varied by data type and was substantially lower for a two-value-per-slot image test than for a grayscale image.

A record claim for data encoding
Moth says its Tessa method has encoded and recovered 1,568 numerical values on an IBM quantum processor, a result the company calls a new record. The preprint’s comparison says the 120-qubit Phoenix experiment encoded and decoded more values than earlier experiments under its stated definition: an arbitrary list of real numbers written to a physical processor and read back.
The PR Newswire release, published on November 16, identifies the processor as an IBM Quantum Nighthawk r2 and says the image was recovered from all 120 qubits. The paper’s acknowledgements describe IBM’s Phoenix as a 120-qubit Nighthawk r2 processor, and its tables list the 1,568-value run on Phoenix, so the two descriptions refer to the same device.
How Tessa encodes data
The preprint by Daniel Bultrini of Moth Quantum was first submitted to arXiv on November 13. It describes Tessa as a hardware-aware method that distributes address and data qubits across a processor’s coupling graph. Nearby data qubits share some address qubits, and the paper says that using the device’s own connections avoids SWAP gates or routing.
The paper reports encoding and decoding up to 1,568 real values using IBM’s Phoenix and Fez quantum computers. In the Phoenix image test, the method placed two values per pixel, representing image value and hue. The comparison includes 128 values recovered from 12 trapped-ion qubits in earlier work; the paper distinguishes those demonstrations from studies that encoded data but read back only a derived quantity, or used a tensor-network approximation rather than an arbitrary list of real values.
Reconstruction quality varied
The Phoenix test encoded 1,568 values using 1,456 CZ gates. The image runs used 8,192 shots per measurement basis. In a separate grayscale test, Phoenix encoded 784 values and Fez encoded 441; their hardware mean absolute errors were 0.130 and 0.131, respectively. The paper reports lower errors after a processing step it calls dynamic-range recovery.
For the two-value-per-slot Phoenix image test, the hardware correlation was 0.69, compared with 0.94 in the paper’s local-noise simulation. The reported value and hue errors were 0.152 and 0.092. Fez encoded 882 values in this mode; its reported hardware errors were 0.144 for value and 0.084 for hue.
Biological sequences and limits
The paper also reports tests involving audio, musical notes, DNA and protein sequences. The biological sequence experiments used 32,768 shots per experiment. Results differed across samples: on Phoenix, MT-CO3 had a 92.3% hardware recovery rate after the best dynamic-range recovery, while the corresponding best-recovery result for SUR1 was 22.6%.
The PR Newswire release says Moth also encoded p53, a protein sequence, and that image-processing Tessa is available to experiment with on the company’s Atlas platform. The paper describes the results as experiments on quantum hardware; its comparison and reconstruction figures depend on the data, encoding mode and recovery measure reported for each test.