Microsoft's three-level quantum computing framework. How to think about where we actually are and what comes next

[Undertaker]

Microsoft Quantum developed a framework laying out three levels of quantum computing progress that is becoming widely referenced for communicating where the field is.

Level 1: NISQ machines. Roughly 1,000 physical qubits, noisy, error-prone, useful for specific research problems and algorithm development but not general commercial computation.

Level 2: Fault-tolerant with a small number of logical qubits. Error correction working reliably, logical qubits with meaningful error rates, able to run specific commercially valuable algorithms end to end.

Level 3: Large-scale fault-tolerant. Millions of physical qubits, thousands of logical qubits, able to run Shor's algorithm at scale for cryptanalysis, simulate molecules at pharmaceutical relevance, and solve optimisation problems that classical supercomputers cannot approach.

In 2026, Quantum Computers Will Reach a New Level

https://spectrum.ieee.org/neutral-atom-quantum-computing

[Bright Hermit]

The three-level framework is useful because it separates the engineering milestones clearly. We are at Level 1 in 2026. The CHIPS Act Anderon foundry investment is aimed at accelerating the transition to Level 2

[Rapid Crossing]

Most quantum advantage claims being made today are at the boundary between Level 1 and Level 2. The Q-CTRL and IBM Fermi-Hubbard result and the UCL turbulence paper are examples of Level 1 hardware producing Level 2-adjacent results for specific problems

[VoidSentinel]

Level 3 is what Q-Day requires. The GQI worst-case three-year estimate for ECC offline attacks is aggressive precisely because reaching Level 3 in three years would be extraordinary

[Northernah]

The framework helps evaluate vendor claims. When a company says quantum advantage ask which level their hardware is at and whether the advantage claim is a Level 1 result on a Level 1 problem or a genuine Level 2 demonstration

[Aisha]

Quantinuum's Helios at 48 logical qubits is the furthest anyone has documented toward Level 2. The Apollo system targeting 2029 is Quantinuum's commitment to a Level 2 machine on a specific timeline