---
title: 'NVIDIA Ising: Open AI Models for Fault-Tolerant Quantum'
description: 'NVIDIA Ising (April 14, 2026): the first open AI model family for quantum computing, with a calibration VLM and decoders that beat pyMatching by 2.5x.'
url: https://nerdleveltech.com/nvidia-ising-open-quantum-ai-models-error-correction
language: en
date: '2026-04-28'
category: ai-ml
tags:
  - nvidia ising
  - quantum computing
  - quantum error correction
  - open-weight model
  - cuda-q
  - nvqlink
  - vision language model
markdown_url: https://nerdleveltech.com/nvidia-ising-open-quantum-ai-models-error-correction.md
attribution: 'Source: NerdLevelTech — https://nerdleveltech.com/nvidia-ising-open-quantum-ai-models-error-correction'
terms: https://nerdleveltech.com/llms.txt
---

# NVIDIA Ising: Open AI Models for Fault-Tolerant Quantum

On April 14, 2026 — World Quantum Day — NVIDIA shipped Ising, an open-source AI model family aimed at two of the hardest problems between today's noisy quantum processors and tomorrow's useful ones: real-time error-correction decoding and processor calibration.[^1][^2] NVIDIA describes Ising as the world's first family of open AI models for quantum computing.[^1] The release pairs a 35B-parameter vision-language model trained on quantum experiment plots with a pair of compact 3D CNN decoders that NVIDIA benchmarks as up to **2.5x faster** and **3x more accurate** than pyMatching, the most widely used open-source decoder for surface-code error correction.[^3][^4]

## TL;DR

Ising is a two-part family. **Ising Calibration** is a Mixture-of-Experts vision-language model — `Ising-Calibration-1-35B-A3B`, 35B total parameters with ~3B active per token across 256 experts (8 routed per token) — built on top of Qwen3.5-35B-A3B with an integrated vision encoder for experiment plots.[^5] On NVIDIA's own QCalEval benchmark (243 entries across 87 scenarios from 22 experiment families) it scores **3.27% above Gemini 3.1 Pro, 9.68% above Claude Opus 4.6, and 14.5% above GPT-5.4**.[^6] **Ising Decoding** is a 3D convolutional neural network that ships in two variants — Fast (0.9M parameters) and Accurate (1.8M parameters) — for real-time surface-code error correction.[^4] At distance d=13 with physical error rate p=0.003, the Fast variant is 2.5x faster and 1.1x more accurate than pyMatching; the Accurate variant is 2.3x faster and 1.5x more accurate. NVIDIA also reports the decoder requires roughly 10x less training data than alternative approaches.[^4][^7] Weights are released under the NVIDIA Open Model License; training frameworks are Apache 2.0 on GitHub; both checkpoints are on Hugging Face.[^8]

## What You'll Learn

- The two model components inside Ising and what each one solves
- Why April 14 — and why "Ising" is the name
- The Calibration MoE architecture and what QCalEval actually measures
- The Decoding 3D-CNN variants and the pyMatching baseline
- How Ising plugs into CUDA-Q QEC and NVQLink for real-time loops
- Licensing, availability, and which labs are already adopting it

## Why April 14 — And Why "Ising"

Two pieces of branding context matter for understanding the launch. First, April 14 is **World Quantum Day**, an initiative quantum scientists started in 2021 to mark Planck's constant — 4.1356677×10⁻¹⁵ eV·s — whose first three rounded digits give the calendar date.[^9] NVIDIA's announcement timed itself to that observance.

Second, the name "Ising" points to the early-twentieth-century Ising model, a lattice model of interacting spins that became one of the foundational simplifications of complex many-body physics.[^1][^2] The metaphor lines up with what these AI models do: take a noisy, high-dimensional measurement signal off a quantum chip, and collapse it into something tractable.

## Ising Calibration: A 35B Vision-Language Model for Qubit Plots

Calibrating a quantum processor is, in practice, hours of staring at plots. Engineers run an experiment, look at a heatmap or a Rabi oscillation, and decide what to tune next. The pipeline is visual, multi-step, and currently human-bound. Ising Calibration is NVIDIA's attempt to automate that loop.

The model is a Mixture-of-Experts vision-language model. The architecture pairs an integrated vision encoder for experimental plot images with the Qwen3.5-35B-A3B MoE language model, giving roughly **35 billion total parameters with ~3 billion active per token** — 256 experts in the MoE layer, 8 routed and 1 shared activated for each token.[^5] The model accepts a plot image plus a structured natural-language question and emits structured technical text.

NVIDIA evaluated it on a benchmark it published alongside the model, **QCalEval**, which it describes as the first vision-language benchmark for quantum calibration plots.[^10]

| QCalEval — what's in it |  |
|---|---|
| Benchmark entries | 243 |
| Scenario types | 87 |
| Experiment families | 22 |
| Plot images (scatter, line, heatmap) | 309 |
| Question types per entry | 6 (1,458 total QA pairs) |
| Hardware coverage | Superconducting qubits + neutral atoms |
| Question categories | Technical description, experimental conclusion, experimental significance, fit-quality assessment, parameter extraction, experiment-success classification |

On QCalEval, Ising Calibration 1 scores higher than the closed frontier models that came up in the same comparison: NVIDIA reports it is **3.27% better on average than Gemini 3.1 Pro, 9.68% better than Claude Opus 4.6, and 14.5% better than GPT-5.4**.[^6] These are NVIDIA-reported numbers on a NVIDIA-published benchmark, so the right framing is "the calibration-tuned model beats general-purpose VLMs at the calibration task it was trained for" rather than "this is a more capable general VLM than GPT-5.4."

## Ising Decoding: Two 3D CNNs Against the pyMatching Baseline

If calibration is the slow human-in-the-loop problem, **decoding** is the fast machine-only one. A real fault-tolerant quantum computer running surface-code error correction has to read syndrome measurements off the chip and infer the most likely set of physical errors — within microseconds, before the next round of gates breaks coherence. Today most groups use pyMatching, an open-source minimum-weight perfect matching decoder, as the standard baseline.[^4]

NVIDIA's Ising Decoding takes a different approach. The decoder is a 3D convolutional neural network that ingests a stacked syndrome volume (two spatial dimensions of the surface code, plus rounds of measurement on the third axis) and emits a correction. Two variants ship at launch:

| Variant | Parameters | vs pyMatching at d=13, p=0.003 |
|---|---|---|
| Decoder SurfaceCode 1 Fast | 0.9M | **2.5x faster latency, 1.1x higher accuracy** |
| Decoder SurfaceCode 1 Accurate | 1.8M | **2.3x faster latency, 1.5x higher accuracy** |

Source: NVIDIA Ising release.[^4][^7]

Two structural things matter beyond the headline numbers. First, because the decoder is a 3D CNN, **the same trained weights run on different surface-code distances and round counts at inference time** — the model is simply applied to a different decoding volume, with no retraining required to change distance or rounds.[^4] Independent third-party work has noted that maintaining the error-correction advantage at substantially larger code distances may require scaling the decoder size proportionally, but small models work across a useful range out of the box.[^20] Second, NVIDIA reports the decoder requires roughly **10x less training data** than alternative approaches at comparable accuracy.[^7]

For deployment, the trained model exports to **ONNX**, then runs through **NVIDIA TensorRT** with **FP8 or FP16** inference on GPU. Optional **INT8 or FP8 post-training quantization** is supported.[^11]

## How It Plugs Into CUDA-Q and NVQLink

Ising is not a standalone artifact — it is the AI half of a quantum-classical loop NVIDIA has been assembling since late 2025.

**NVQLink**, announced October 28, 2025 at GTC Washington D.C., is the GPU-to-QPU interconnect: an open system architecture that uses RDMA over Converged Ethernet (RoCE) on standard NVIDIA networking hardware to deliver deterministic communication between quantum control electronics and GPU servers, with a maximum round-trip latency of 3.96 microseconds.[^12] It is the "wire" between the quantum chip and the AI decoder.

**CUDA-Q QEC 0.6**, released the same day as Ising, is the software side. It introduces two new real-time-capable decoder pipelines over NVQLink: **RelayBP**, a GPU-accelerated belief-propagation decoder for quantum LDPC codes, and the **NVIDIA Ising pre-decoder** combined with **PyMatching** as a global decoder for the surface code.[^11][^13] The pre-decoder pattern — neural net does the fast, high-accuracy first pass, classical algorithm cleans up residual cases — is how Ising drops into a working error-correction loop without replacing the existing pipeline wholesale.

Why this matters: the entire round-trip latency, from qubit measurement to corrective gate, has to fit inside the coherence window of the underlying qubits. For superconducting qubits operating on tens-of-nanoseconds gate times and surface-code cycle times of around 1 microsecond, decoders need to keep up in real time — the closer to the cycle time, the better.[^14][^16] A decoder-only speedup of 2.5x is therefore not a benchmark trophy; it is what closes the gap between offline analysis and a working real-time QEC loop.

## The Wider Quantum Decoder Race

Ising lands in a busy field. In March 2026, **QpiAI** demonstrated a union-find decoder on a 64-qubit Kaveri processor that completed each error-correction cycle in roughly **1.5 microseconds**, fast enough to run inside the coherence window of superconducting qubits.[^15] **Google's Willow** processor — disclosed in late 2024 and refined since — runs a real-time decoder on a 101-qubit distance-7 surface code, achieving **0.143% ± 0.003% error per cycle** and beating its best physical qubit's lifetime by a factor of **2.4 ± 0.3**.[^16] **AlphaQubit**, a Google-DeepMind transformer-based decoder published in *Nature* in November 2024, was demonstrated on Google's Sycamore processor at distance-3 and distance-5 surface codes, with simulated runs out to distance 11.[^17] A separate convolutional neural-network decoder (Gicev et al., 2023, *Quantum*) has been demonstrated on simulated surface codes with code distances exceeding 1000 — over 4 million physical qubits — the largest ML-based surface-code decoder demonstration to date.[^19]

These results are not all directly comparable — Willow's 0.143% per cycle is a full-stack hardware-plus-decoder logical error rate, QpiAI's 1.5 µs is a hardware-decoder cycle time on a single Kaveri device, and NVIDIA's 2.5x speedup is a decoder-only benchmark against pyMatching. What Ising adds to the picture is a different kind of contribution: **open weights, an open training framework, an open benchmark, and a deployment pipeline that any research group with NVIDIA GPUs can adopt**.

## Licensing and Availability

| Component | Where | License |
|---|---|---|
| Ising Calibration 1 weights (35B-A3B) | Hugging Face, NVIDIA NIM, NVIDIA Build | NVIDIA Open Model License |
| Ising Decoder SurfaceCode 1 weights (Fast + Accurate) | Hugging Face | NVIDIA Open Model License |
| Ising training framework | GitHub: NVIDIA/Ising-Decoding | Apache 2.0 |
| QCalEval benchmark dataset | Hugging Face datasets, GitHub: NVIDIA/QCalEval | Open access |
| CUDA-Q QEC 0.6 | NVIDIA developer site | Existing CUDA-Q license |

Source: NVIDIA Ising release notes and Hugging Face model cards.[^8][^11]

The weights/code split — model weights under NVIDIA's Open Model License, training framework under Apache 2.0 — is the same pattern NVIDIA used for Nemotron and recent NIM-distributed models. It gives QPU operators latitude to keep proprietary calibration data on-prem while still letting them fine-tune the published checkpoints for their specific hardware.

## Early Adopters

NVIDIA's launch press release lists two separate adopter tracks. For **Ising Calibration**, it names 12 organizations: **Academia Sinica**, **Atom Computing**, **Conductor Quantum**, **EeroQ**, **Fermi National Accelerator Laboratory**, **Harvard John A. Paulson School of Engineering and Applied Sciences**, **Infleqtion**, **IonQ**, **IQM Quantum Computers**, **Lawrence Berkeley National Laboratory's Advanced Quantum Testbed**, **Q-CTRL**, and the **U.K. National Physical Laboratory (NPL)**.[^1] For **Ising Decoding**, it names a largely different set: **Cornell University**, **EdenCode**, **Infleqtion**, **IQM Quantum Computers**, **Quantum Elements**, **Sandia National Laboratories**, **SEEQC**, **UC San Diego**, **UC Santa Barbara**, **University of Chicago**, **USC**, and **Yonsei University**.[^1] The mix spans both hardware modalities the QCalEval benchmark covers, with **Infleqtion** on neutral atoms and **IQM** on superconducting qubits — both companies also appear on the decoding-adopter list.

## FAQ

**Is Ising a quantum AI model — meaning, does it run on quantum hardware?**

No. Ising is a classical AI model family that runs on NVIDIA GPUs. It is built **for** quantum computing — its job is to help calibrate and error-correct quantum processors — but the inference happens classically.

**Does Ising replace pyMatching?**

Not quite. NVIDIA explicitly positions the Ising decoder as a **pre-decoder** in CUDA-Q QEC 0.6, paired with PyMatching for the surface code path.[^13] The neural net does the bulk of the work fast and accurately; the classical decoder cleans up the residual cases. You can also run Ising standalone if you prefer.

**What is QCalEval and how independent is it?**

QCalEval is the first vision-language benchmark for quantum calibration plots: 243 entries, 87 scenario types, 22 experiment families, 6 question categories per entry.[^10] It was authored by NVIDIA Research and published with the model, so it is not an independent benchmark — but the dataset and scoring scripts are public on Hugging Face and GitHub, which means independent groups can reproduce the numbers and report on other models.[^18]

**Can I run Ising Calibration locally?**

Yes, in principle. The full checkpoint is `Ising-Calibration-1-35B-A3B` on Hugging Face. The "A3B" suffix is the giveaway: only ~3B parameters are active per token at inference, so you get the inference cost of a small dense model with the parameter breadth of a 35B MoE. Realistically you still want a multi-GPU box or an H100/H200/B200-class accelerator to host the full weights at usable latency.

**How does this compare to AlphaQubit?**

Both are neural-network decoders for the surface code, but they're optimizing for different points on the design space. AlphaQubit is a recurrent transformer demonstrated on Google's Sycamore at distance-3 and distance-5 surface codes (with simulated runs out to distance 11) and beats matching decoders on accuracy.[^17] Ising Decoding is a compact 3D CNN (sub-2M parameters) targeted at real-time deployment over NVQLink, with an explicit FP8/TensorRT inference path. They are complementary in research, but Ising is the one with an end-to-end open deployment story today.

## Bottom Line

NVIDIA Ising is the first time a major hyperscaler has shipped open weights, an open training framework, and an open benchmark targeted specifically at the calibration and error-correction subproblems of quantum computing — paired with a real-time deployment pipeline (NVQLink + CUDA-Q QEC) that lets the AI half of the loop actually meet the microsecond budget the physics demands. Whether the **2.5x latency** and **3x accuracy** numbers hold up under independent replication on third-party QPUs is the question to watch over the next two quarters. But the architectural bet — that the path to useful quantum computers runs through AI-accelerated classical co-processors, not just better qubits — is now openly testable by anyone with an NVIDIA GPU and a quantum testbed.

For broader context on AI in scientific computing, see our earlier coverage of [thermodynamic computing as a post-transistor frontier](thermodynamic-computing-the-next-leap-beyond-transistors) and the [DeepSeek V4 open-source frontier release](deepseek-v4-open-source-frontier-million-token-context).

[^1]: NVIDIA Newsroom, "NVIDIA Launches Ising, the World's First Open AI Models to Accelerate the Path to Useful Quantum Computers," April 14, 2026. https://nvidianews.nvidia.com/news/nvidia-launches-ising-the-worlds-first-open-ai-models-to-accelerate-the-path-to-useful-quantum-computers
[^2]: The Quantum Insider, "NVIDIA Launches Ising, the World's First Open AI Models to Accelerate The Path to Useful Quantum Computers," April 14, 2026. https://thequantuminsider.com/2026/04/14/nvidia-launches-ising-the-worlds-first-open-ai-models-to-accelerate-the-path-to-useful-quantum-computers/
[^3]: Tom's Hardware, "Nvidia releases open AI models for quantum computing tasks — 'Ising' said to be 2.5x faster and 3x more accurate than existing tools for decoding," April 2026. https://www.tomshardware.com/tech-industry/artificial-intelligence/nvidia-releases-ising-open-ai-models
[^4]: NVIDIA Developer Blog, "NVIDIA Ising Introduces AI-Powered Workflows to Build Fault-Tolerant Quantum Systems," April 2026. https://developer.nvidia.com/blog/nvidia-ising-introduces-ai-powered-workflows-to-build-fault-tolerant-quantum-systems/
[^5]: Hugging Face, model card "nvidia/Ising-Calibration-1-35B-A3B," 2026. https://huggingface.co/nvidia/Ising-Calibration-1-35B-A3B
[^6]: NVIDIA Build, "ising-calibration-1-35b-a3b Model by NVIDIA — model card," 2026. https://build.nvidia.com/nvidia/ising-calibration-1-35b-a3b/modelcard
[^7]: MarkTechPost, "NVIDIA Releases Ising: the First Open Quantum AI Model Family for Hybrid Quantum-Classical Systems," April 19, 2026. https://www.marktechpost.com/2026/04/19/nvidia-releases-ising/
[^8]: GitHub, "NVIDIA/Ising-Decoding: A training framework for AI Quantum Error Correction Decoders." https://github.com/NVIDIA/Ising-Decoding
[^9]: World Quantum Day, "Why April 14." https://worldquantumday.org/why-april-14
[^10]: NVIDIA Research, "QCalEval: Benchmarking Vision-Language Models for Quantum Calibration Plot Understanding," April 2026. https://research.nvidia.com/publication/2026-04_qcaleval-benchmarking-vision-language-models-quantum-calibration-plot
[^11]: NVIDIA Quantum, "CUDA-Q QEC 0.6 Enables Real-Time QEC with NVQLink," April 14, 2026. https://nvidia.github.io/cuda-quantum/blogs/blog/2026/04/14/cudaq-qec-0.6/
[^12]: NVIDIA Developer Blog, "NVIDIA NVQLink Architecture Integrates Accelerated Computing with Quantum Processors." https://developer.nvidia.com/blog/nvidia-nvqlink-architecture-integrates-accelerated-computing-with-quantum-processors/
[^13]: NVIDIA, "Open AI Models for Quantum Computing | NVIDIA Ising." https://www.nvidia.com/en-us/solutions/quantum-computing/ising/
[^14]: arXiv, "Demonstrating real-time and low-latency quantum error correction with superconducting qubits," 2410.05202, 2024. https://arxiv.org/html/2410.05202v1
[^15]: The Quantum Insider, "QpiAI Achieves High-Speed Quantum Error Correction on Superconducting Systems with New Decoder Platform," March 25, 2026. https://thequantuminsider.com/2026/03/25/qpiai-high-speed-quantum-error-correction-decoder/
[^16]: Nature, "Quantum error correction below the surface code threshold," 2024. https://www.nature.com/articles/s41586-024-08449-y
[^17]: Nature, "Learning high-accuracy error decoding for quantum processors" (AlphaQubit), 2024. https://www.nature.com/articles/s41586-024-08148-8
[^18]: Hugging Face Datasets, "nvidia/QCalEval," 2026. https://huggingface.co/datasets/nvidia/QCalEval
[^19]: Gicev et al., "A scalable and fast artificial neural network syndrome decoder for surface codes," *Quantum*, July 2023. https://quantum-journal.org/papers/q-2023-07-12-1058/
[^20]: Quantum Computing Report, "Evaluating Neural Pre-Decoding with NVIDIA Ising: From Surface to Bivariate Bicycle Codes" — third-party analysis noting that decoder size must scale with code distance for sustained error-correction advantage, 2026. https://quantumcomputingreport.com/evaluating-neural-pre-decoding-with-nvidia-ising-from-surface-to-bivariate-bicycle-codes/