Describe the task
Start with the circuit, algorithm, or behavior you want. GrayGate interprets the task and its constraints.
GRAY AREA LABS / QUANTUM SOFTWARE Private beta · request access
Quantum development.
Grounded in evidence.
Gray Area Labs builds software for quantum developers. GrayGate turns your request into AI-generated Qiskit code, then uses simulation and verification checks to help you review the result.
OUR PLATFORM
Meet GrayGate.
Inspect the code, checks, and run details below.
This is an interactive example, not a live run.
GrayGate /Quantum workspaceILLUSTRATIVE EXAMPLECreate a Bell state
Create a two-qubit Bell state circuit and measure both qubits.
Qiskitbell_state.py
from qiskit import QuantumCircuit
from qiskit_aer import AerSimulator
# Prepare an entangled pair
circuit = QuantumCircuit(2)
circuit.h(0)
circuit.cx(0, 1)
circuit.measure_all()
# Simulate and inspect the measurement counts
simulator = AerSimulator()
result = simulator.run(circuit, shots=1024).result()
print(result.get_counts())Python · QiskitExample source · ready to copy
EXAMPLE VERIFICATION REPORT
Three checks. Inspectable evidence.
- Static validation
Python syntax and expected imports are checked before execution.
- Simulation
The illustrated output contains the correlated outcomes 00 and 11.
- Reference comparison
The unmeasured state is compared with the expected Bell state.
Illustrative checks only. This page does not execute or verify code.
EXAMPLE RUN CONFIGURATION
A small circuit, made explicit.
- Task
- Two-qubit Bell state
- Simulator
- Qiskit Aer
- Shots
- 1,024
- Expected outcomes
- 00 and 11, approximately 50% each
- Data source
- Illustrative example
No model request or quantum execution is made by this example.
Qiskit is an open-source framework for quantum computing.BUILT FOR 
Qiskit
QiskitTHE WORKFLOW
More than a code response.
A path from intent to evidence.
With room to inspect, refine, and try again.
Generate with context
Relevant Qiskit documentation and reference examples inform the generated program.
Execute and verify
Simulation and independent checks surface failures. Repair loops work through the findings.
Review the evidence
Inspect the code, execution results, and verification signals before deciding what comes next.
Inside the verification pipeline
Task intake and contract definition → version-aware retrieval → code generation → static checks → sandboxed execution → reference, generated-test, and quantum-property checks → confidence assessment → repair and finalization. Checks vary by task; a passing result is evidence, not a guarantee of correctness.
MEASURED, NOT ASSUMEDPut the code
Explore the full leaderboard
Put the code
to the test.
Quantum programs need more than plausible syntax. Here’s how GrayGate and the recorded model runs performed on 151 quantum programming tasks.
91.6%
GrayGate mean pass rate3 recorded pipeline runs · see methodology below
151tasks per run
19recorded runs
7model / pipeline entries
BENCHMARK OVERVIEW
Historical snapshotQiskit-HumanEval-Hard
GrayGateVerification pipeline
91.6%Gemini 3 ProPreview
59.9%Gemini 3 FlashPreview
55.6%GPT-5.2Direct model
46.1%Qwen 3.5397B A17B
38.4%DeepSeek ReasonerDirect model
32.5%Qiskit Qwen 2.514B · local model
25.8%2026-02-17 – 2026-03-03Download source data ↓
How to read these results
Metric. Mean of the task pass percentage for every completed 151-task run per model in the saved dataset. Bars share a 0–100% scale. Ranges show the lowest and highest recorded run, not confidence intervals.
Context. Historical, internally recorded results; not independently reproduced. Configurations and Qiskit versions differ. GrayGate uses a repair pipeline; these are not equal-budget single-attempt comparisons. Dataset revision and GrayGate base model are not recorded in these run rows.
Source. Snapshot of the repository’s benchmark database, completed 2026-02-17 through 2026-03-03. The downloadable data includes run IDs, counts, available settings, and a source checksum. The full leaderboard may change as new runs are added.
UNDER THE HOODOne answer.
One answer.
Multiple ways to check it.
A program can run and still be wrong.
GrayGate brings different signals together
so you can see more than the final output.
INSIDE GRAYGATE
Illustrated workflowA closer look at the evidence.
bell_state.pyGenerated code + task contract
INPUT2 qubitsPrepare → entangle → measure
01
Static analysis
Check the structure.
- Python syntax
- Imports & contracts
02
Execution
Observe the behavior.
- Isolated subprocess
- Simulation results
03
Verification
Compare the evidence.
- Reference comparisons
- Tests & properties
An inspectable resultCode · execution outcomes · confidence signals
REVIEWSomething doesn’t hold up? Findings guide the repair loop.
A conceptual view of the pipeline. Checks depend on the task and configuration.
Context before generation
Qiskit documentation and reference examples ground the request. Task contracts give the pipeline explicit requirements to check.
Evidence beyond execution
Reference comparisons, generated tests, and quantum-property checks examine different aspects of the result. A successful execution is one signal among several.
Repair with a reason
Errors and low-confidence findings guide another attempt. Review the resulting code and evidence, including the checks that did not pass.
FROM QUESTION TO CIRCUIT
What will you explore?
Start with a concrete task.
Keep the implementation and evidence in view.
RESEARCH & PROTOTYPING
Prototype quantum circuits.
Move from a circuit description to code you can read, simulate, and refine. Compare the intended behavior with the observed result.
EXAMPLE QUESTION
Request beta access →“Create a three-qubit GHZ state and measure each qubit.”
QUANTUM DEVELOPMENT
Develop Qiskit programs.
Describe the requirements, follow generation and repair progress, and inspect the final code before bringing it into your project.
EXAMPLE QUESTION
Request beta access →“Build a quantum Fourier transform circuit for four qubits.”
LEARNING & TEACHING
Learn through examples.
Explore how a quantum operation is expressed in Qiskit. Use the code and simulation evidence as a starting point for your own investigation.
EXAMPLE QUESTION
Request beta access →“Create a Bell pair and show the measurement outcomes.”
A FEW GOOD QUESTIONSBefore you
Before you
start building.
More about the product,
the checks, and getting access.
What is the relationship between Gray Area Labs and GrayGate?
Gray Area Labs builds the software. GrayGate is our quantum code-generation and verification platform, focused on Qiskit workflows.
Does GrayGate run my code?
The generation pipeline can execute code in an isolated subprocess and use simulation results as part of its checks. The example on this landing page is illustrative; it does not submit a model request or execute quantum code.
Does passing verification guarantee correctness?
No. Verification provides evidence within the available checks and task constraints. Review the code, assumptions, and results, and test it in your own workflow. Some tasks have stronger reference checks than others.
Do I need access to a quantum computer?
The illustrated workflow uses Qiskit Aer simulation, so it does not require access to physical quantum hardware. Hardware execution is separate from the simulation and verification shown here.
How do I get started?
GrayGate is in private beta. Request access and tell us about your work. Existing users can sign in through the same page. We’re interested in feedback from researchers, developers, and educators.
GRAY AREA LABS
About the company.
We build quantum development tools, starting with GrayGate. We’re working with researchers, developers, and educators to improve how quantum code is generated and checked.
Wyatt Greene
Founder · Longmont, ColoradoBuilding GrayGate’s verification engine and developer experience.
GET STARTED WITH GRAYGATEYour next quantum program
Request early access