OpenAI Math Preprint Shows Quantum Search Can Beat Classical By Nearly Fourth Power

The short version

OpenAI researchers posted a proof that quantum computers can need far fewer queries than classical randomized algorithms on some total Boolean functions. The gap is nearly quartic, which rules out the old hope that classical cost might always grow only as the cube of quantum cost.

What the researchers did

The note A Nearly Quartic Separation Between Randomized and Quantum Query Complexity is dated 5 October 2026 in OpenAI’s public math preprint folder. It defines total Boolean functions where bounded-error randomized query complexity R(f) and quantum query complexity Q(f) diverge sharply. Both models count worst-case bit queries with error at most one third on every input.

What they found

The authors build examples where no universal bound of the form R(f) = O((1 + Q(f))^α) holds for any α below 4. That matches the known quartic upper bound’s exponent and disproves a widely hoped cubic ceiling. In plain terms, there are problems where quantum search wins by almost the fourth power gap that theory already allowed as a ceiling.

Why it matters for you

If you fund quantum algorithms or set cryptography timelines, query complexity is a clean yardstick for when quantum advantage is real on structured search, not just hype. This result says the exponent 4 in upper bounds is tight, so planners should not assume milder classical catch-up.

What to keep in mind

This is pure complexity theory, not a hardware benchmark on IBM or Google chips. Preprint from OpenAI math repository, not yet peer-reviewed in a journal. Functions are constructed for separation, not for immediate commercial use.

Query complexity counts how many input bits an algorithm must read. Randomized classical algorithms may flip coins while reading. Quantum algorithms may use superposition and interference. A total Boolean function maps every input length to zero or one. Bounded error means the algorithm can be wrong on at most a third of inputs. Worst case means the count applies to the hardest input.

The open question was how fast classical cost grows when quantum cost is small. Some hoped classical cost might grow only as the cube of quantum cost. This paper builds functions where that hope fails. The gap is close to the fourth power in the exponent. That matches the best known upper limit in theory. Matching an upper limit tells researchers the ceiling is tight, not loose.

This is not a claim that your laptop search is obsolete tomorrow. It guides which separations are possible in proof systems. Cryptography teams track these limits when judging quantum threats to structured search problems. Always treat the PDF as the source for exact function names and constants.

Quantum query models count questions to an oracle. Classical randomized models may flip coins. Quantum models may ask superposed questions. A separation proves a gap is possible. It does not ship a product. Theory papers still guide funding and education. Students learn which exponents are tight. OpenAI posted the PDF in its math repo. Download the PDF for theorem numbers. We do not restate proofs here. We focus on why planners should care.

Teams share papers so others can repeat the work. A preprint is an early public draft. Peer review may change claims later. Numbers in tables can move after feedback. Always read the linked paper for exact values. This desk note tracks when we last checked sources. We remove claims if sources disagree. If you build on this work, cite the primary paper. If you spot an error, tell the publisher through the site contact path.

Key takeaways

• Randomized and quantum query costs can separate by nearly a quartic factor.
• The result kills α < 4 universal power laws linking R(f) and Q(f).
• Exponent 4 in known upper bounds is optimal in the worst case sense shown.
• Posted 5 October 2026 alongside other OpenAI math releases.
• Useful for algorithm designers and post-quantum roadmaps, not chip buyers alone.

Photo: Pexels

Sources

• Primary — OpenAI math preprints, A Nearly Quartic Separation Between Randomized and Quantum Query Complexity (5 October 2026). Theorem statement and exponent tightness.