The Open Quantum Challenge: Quantum Simulation and QEC Decoding on Classical GPUs

Abstract

The Open Quantum Challenge is an open competition that reframes two of the hardest problems in quantum computing, state simulation and quantum error correction (QEC), as reproducible tasks on classical hardware. Participants compete using only a GPU and a publicly provided evaluation harness. No access to quantum hardware is required. The challenge runs in two concurrent seasons with a total prize pool of 2,000 USD.

• Challenge Space: https://huggingface.co/spaces/FINAL-Bench/open-quantum-challenge

Motivation

Progress in quantum computing is gated by hardware access. Devices are scarce, queues are long, and the cost of experimentation is high. Yet a large and important fraction of quantum research is classical in nature. Validating a simulation method, stress-testing a decoder, and benchmarking algorithms are all computations that modern GPUs perform well.

The Open Quantum Challenge is built on that observation. By moving the work onto classical accelerators and standardizing the evaluation, it lowers the barrier to entry and produces a shared, reproducible leaderboard rather than a scatter of private, non-comparable results.

The two seasons

Season 1: Quantum simulation

Participants implement methods that reproduce the behavior of a target quantum system as accurately as possible on a GPU. Submissions are scored against a defined ground truth, so the measurement is objective and reproducible.

Season 2: QEC decoder

Participants build a decoder for quantum error correction: given syndrome information from coded quantum states, the decoder must infer and correct the underlying errors. Decoder quality is a direct, well-posed metric, which makes it an excellent benchmark target.

Both seasons open on the same day and run in parallel. A participant may enter either or both.

Why these tasks make good benchmarks

Both tasks have well-defined ground truth. Unlike open-ended generation, where scoring requires fuzzy matching or a judge model, quantum simulation and QEC decoding are scored against known targets. That yields three properties that matter for a serious benchmark:

1. Objectivity. Submissions are compared to a defined reference, not a subjective rater.
2. Reproducibility. The harness is public, so any result can be reproduced independently.
3. Comparability. Every entrant is measured the same way, which makes the leaderboard meaningful.

How to participate

The on-ramp is deliberately short:

1. Open the challenge Space: https://huggingface.co/spaces/FINAL-Bench/open-quantum-challenge
2. Copy the participation guide from the page.
3. Paste it into an agentic coding environment such as Codex or Claude Code, which will scaffold your entry.
4. Run against the public harness and submit.

Because the guide is designed to be handed directly to a coding agent, a participant can move from first look to a working submission in a single session, without specialized quantum tooling.

Prize

The total prize pool is 2,000 USD across the challenge.

Who should enter

The challenge is open to anyone with a GPU and an interest in quantum computing: ML engineers curious about quantum methods, researchers who want a reproducible venue for decoder and simulation work, and students looking for a well-scoped, high-signal project. No prior quantum-hardware experience is assumed.

Closing

Quantum error correction and simulation will shape whether and when quantum computing becomes useful. Opening both to the much larger community of people who own a GPU is a practical way to accelerate that work and to measure it honestly. We invite you to take part.

• Challenge: https://huggingface.co/spaces/FINAL-Bench/open-quantum-challenge
• Organizer: FINAL-Bench (https://huggingface.co/FINAL-Bench), built by VIDRAFT