🚀 Announcing the TensorCircuit-NG Whitepaper: The Next-Gen Universal Platform for Quantum Computing & Simulation!

[refraction-ray]

Hello everyone!

We are beyond excited to announce the release of our new whitepaper: TensorCircuit-NG: A Universal, Composable, and Scalable Platform for Quantum Computing and Quantum Simulation.

TensorCircuit-NG (tcng) represents a massive leap forward. It transitions the original TensorCircuit from a fast quantum circuit simulator into a comprehensive, AI-native quantum software ecosystem designed to bridge the gap between quantum physics, artificial intelligence, and high-performance computing (HPC).

Here is a quick overview of the new features, the advantages of TCNG, and why we believe it is the one-for-all choice for modern quantum research.

---

✨ What's New: Key Features

TensorCircuit-NG moves beyond traditional state-vector updates by establishing a unified, tensor-native programming paradigm. It fuses quantum circuits, tensor networks, and neural networks into a single, end-to-end differentiable computational graph.

• Backend-Agnostic & ML-Integrated: Built natively on top of industry-standard machine learning frameworks (JAX, TensorFlow, PyTorch).
• Advanced Hardware Acceleration: Full, native support for Automatic Differentiation (AD), Just-In-Time (JIT) compilation, and Vectorized Parallelism (VMAP) to remove Python loop overheads.
• Expanded Physics Capabilities: Comprehensive out-of-the-box support for:
  • Approximate circuit simulation (MPS circuits)
  • Analog dynamics and continuous-time many-body dynamics
  • Fermion Gaussian states
  • Qudit systems
  • Scalable, high-level customized noise modeling
• Distributed Computing at Scale: To tackle the exponential complexity of deep quantum circuits, TCNG implements advanced distributed computing strategies, including automated data parallelism and model-parallel tensor network slicing.
• AI-Native Workflows: Purpose-built for agentic research and automated scientific discovery with built-in rules and agentic skills.

⚡ The Advantages of TCNG

• Unprecedented Scalability & Performance: We have validated TCNG on multi-node GPU clusters, demonstrating near-linear speedups for distributed Variational Quantum Algorithms (VQAs).
• End-to-End Differentiability: Because physics logic is decoupled from the computational backend, you get seamless gradients. You can optimize tensor network states for many-body physics, hybrid QML architectures, or analog pulse controls without manual gradient derivation.
• Flagship Application Support: TCNG unlocks massive simulations that were previously intractable, such as end-to-end Quantum Machine Learning (QML) for CIFAR-100 computer vision and efficient pipelines translating quantum states to classical neural networks via classical shadows.

🥇 Why it is the "All-in-One" Solution in Quantum Research

The frontier of quantum research has expanded significantly. Researchers no longer just need to simulate logic gates—they need to optimize hybrid models, simulate noise, study continuous physics, and run on supercomputers.

TensorCircuit-NG is built to be the unified home for all these needs. Whether you are:

• A Condensed Matter Physicist modeling continuous-time many-body dynamics or exploring tensor network states,
• A Control Engineer optimizing analog pulses or studying the effects of complex noise models,
• An ML Practitioner embedding parameterized quantum layers into complex, distributed classical pipelines...

...TensorCircuit-NG fuses digital logic, analog dynamics, tensor networks, and neural networks under one roof. It provides a seamless transition from writing small, proof-of-concept algorithms on your laptop to deploying large-scale hybrid algorithms and many-body physics models on GPU supercomputers.

---

Read the full whitepaper here: https://arxiv.org/html/2602.14167v1

We encourage you to check out the paper, explore the 140+ new example scripts in the repo, and try it out. Let us know what you think below, what features you're most excited about, and how you plan to use tensorcircuit-ng in your next project! 🌌⚛️