Papex

Problem · Mathematics · Under review

Navier–Stokes existence and smoothness

Do smooth solutions of the 3D Navier–Stokes equations always exist?

Summary

Asks whether smooth, globally defined solutions always exist for the three-dimensional equations of fluid flow, or whether they can break down.

AI-assisted claims

Proof claimed · Revised · 08 Sep 2026

On the Navier–Stokes Millennium Prize Problem

“Our system produced an analytical proof and a Lean formalization that an initially smooth fluid at rest can develop a singularity in a finite time. The fluid has a smooth force applied to it, and its energy remains finite through the entire dynamics, from rest to the formation of the singularity. This resolves the Navier–Stokes Millennium Prize problem by establishing statement “C” (and also “D”) in the official Millennium Prize formulation.”

By
OpenAI
AI involvement
We used a system of coordinating agents powered by our internal model. The agents had access to tools such as the ability to read from a cached version of the internet and the ability to run code. … The groups varied in size, and the group that produced the Navier–Stokes resolution involved on the order of 10,000 concurrent agents.
Model
an internal model that is significantly more capable than GPT-6 Astra, as stated
Where
OpenAI website (research post), with paper PDF and Lean repository
Formal proof
Lean 4, published by the source
Repository
openai/NavierStokesAndEuler at f9e8bc5 · read 21 Nov 2026
Source last read
14 Oct 2026

Checks

1 exposition · 1 objection · 1 status-holder statement · 1 author update · 4 responses

  • 10 Sep 2026 · Author revision · OpenAI: “Update — September 10, 2026: We have updated “Concurrent work” with findings from our investigation into whether user inputs could have influenced this result.” Source
  • 11 Sep 2026 · Status holder · Clay Mathematics Institute: Statement by the body that administers the Millennium Prize, on the announcement: it names no evaluation outcome and says the prize rules govern evaluation and credit. “The rules governing the prizes describe the process for evaluating what has been achieved and for assigning credit. The process is deliberately unhurried, but we will provide updates.” Source
  • 12 Sep 2026 · Response · Silvia De Toffoli and Eamon Duede: “What OpenAI has given us is an answer. But it is not clear that they have delivered a fruitful solution. Perhaps, we will find that they have, but at the moment, the situation is far from clear.” Source
  • 23 Sep 2026 · Response · Tapio Schneider: “The claimed result says that a fluid starting from rest, driven by just the right smooth stirring, develops an infinite velocity spike in finite time while its total kinetic energy remains bounded; the forcing is constructed to sustain the collapse. (The unforced version of the problem, whether smooth solutions exist for all time without external forcing, remains open.)” Source
  • 28 Sep 2026 · Exposition · Zhen Lei, Xiao Ren (arXiv:2609.35406): Expository re-presentation of the profile-construction part of OpenAI's manuscript only; the cancellation by oscillatory pulses is left to a companion Part II. The authors call the work a major advance but do not state that they checked the whole proof. “This paper gives a readable and accessible version of the profile-construction part of OpenAI's manuscript "Finite Time Blowup for Navier-Stokes". We regard OpenAI's work as a major advance on the Navier-Stokes Millennium Prize Problem.” Source
  • 04 Oct 2026 · Response · Diego Córdoba and Luis Martínez-Zoroa: “OpenAI’s September 8, 2026 announcement concerns two separate claims: blow-up for ordinary Navier–Stokes with smooth forcing, and blow-up for unforced Euler from smooth initial data. The first, if verified under the stated hypotheses, would establish an allowed breakdown alternative in the Millennium Problem formulation. It is distinct from unforced Navier–Stokes blow-up. At this moment we are not in a position to say anything substantive about a comparison between OpenAI’s constructions and our cascade methods.” Source
  • 06 Oct 2026 · Objection · Alexander Bastounis, Fabian Circelli, Anders C. Hansen (arXiv:2610.08144): Correspondence between OpenAI's Lean formalization and its natural-language proof of blow-up; the paper reports that the two do not match. “In particular, we show that the formalised Lean proof does not correspond to the NL proof of blow-up of solutions to the Navier-Stokes equations.” Source
  • 09 Oct 2026 · Response · Thomas Hales: “The Navier-Stokes formalization has already been confirmed by more than a dozen proof-checkers. Cross-checking the proof by different kernels does not remove all doubt.” Source

Claims as their sources state them. Listing a claim is not a check by Papex; the status stays the source’s.

All AI-assisted claims

Why it is open

The equations are Newton's law for a small element of fluid, acted on by pressure, internal friction (viscosity) and any external force, in a fluid that cannot be compressed [4]. Smooth solutions are known to exist for a short time, and for all time when the initial velocity is small; otherwise, either solutions stay smooth for ever or some smooth initial velocity reaches a finite "blowup time", at which the velocity becomes unbounded [4]. The two-dimensional case was settled long ago, but Charles Fefferman notes that it gives no hint about three dimensions [4].

The prize problem asks for a proof of one of four statements [4]. Statements (A) and (B) say that, with no external force, every smooth initial velocity leads to a solution that stays smooth for all time, in all of space (with bounded energy) or in a periodic box. Statements (C) and (D) say that breakdown can happen: for some smooth initial velocity and some smooth external force with the stated decay or periodicity, no such solution exists. The same questions are "also open and very important" for the Euler equations, the case of zero viscosity, which are not on the prize list [4]. Under the prize rules, the Clay Mathematics Institute (CMI) considers a proposed solution only after it has appeared in a qualifying outlet, at least two years have passed, and it has general acceptance in the mathematics community [11].

Fefferman wrote that "standard methods from PDE appear inadequate to settle the problem" [4]. CMI's problem page says that mathematicians and physicists believe turbulence can be explained and predicted through an understanding of these solutions [11]. In September 2026 a proof of statements (C) and (D) was announced (see "Recent claims") [10].

What is known

Weak solutions (1934–2022). In 1934 Jean Leray showed that the three-dimensional equations always have a weak solution, one that satisfies them in an averaged sense [1, 4]. Fefferman's description records that uniqueness of weak solutions was not known and that attempts to prove them smooth had only partial success [4]. Luis Caffarelli, Robert Kohn and Louis Nirenberg proved in 1982 that the singular set of a suitable weak solution has zero one-dimensional parabolic Hausdorff measure, so it cannot contain a curve of the form x = φ(t); Fefferman called this the best partial regularity theorem known when he wrote [2, 4]. In 2003 Luis Escauriaza, Gregory Seregin and Vladimír Šverák proved that unforced solutions whose L³ norm in space, a scale-invariant quantity, stays bounded in time are smooth [3, 10]. In 2022 Dallas Albritton, Elia Brué and Maria Colombo exhibited two distinct Leray solutions with zero initial velocity and the same force; their solutions "live precisely on the borderline of the known well-posedness theory" [6].

Blow-up for related equations (2016–2026). In 2016 Terence Tao proved finite-time blow-up for an "averaged" Navier–Stokes equation with the same energy identity, showing that a proof of global regularity must use finer structure of the nonlinear term than harmonic-analysis estimates and the energy identity; he also proposed a programme for adapting such blow-up to the true equations [5]. For the Euler equations, Jiajie Chen and Thomas Hou review a computer-assisted proof of nearly self-similar blow-up for the 2D Boussinesq and 3D axisymmetric Euler equations in a smooth bounded domain, noting that the general problem "remains unresolved" [7]. A 2025 preprint by Yongji Wang and colleagues at Google DeepMind and several universities reports unstable self-similar solutions of the incompressible porous media equation and of 3D Euler with boundary, some at a precision that "meets the requirements for rigorous mathematical validation via computer-assisted proofs" [8]. In September 2026 Tao described preprints by Levent Alpöge and Tristan Buckmaster, building on Diego Córdoba and Luis Martínez-Zoroa, giving blow-up with a smooth force for the porous media, 2D Boussinesq and 3D Euler equations; they "do not quite achieve" Navier–Stokes blow-up, but Tao judged that their method "has a high likelihood of also extending to Navier-Stokes" [9]. He added that "it is now widely expected" that the Navier–Stokes equations can develop singularities from smooth data and a smooth force, "and it should even be possible to do without the forcing term"; a blog post records opinion, not proof [9]. Córdoba, Martínez-Zoroa and Fan Zheng have proved forced blow-up when viscosity is replaced by a fractional dissipation of order below about 0.093, against order 2 for ordinary viscosity, a gap they call "substantial" [13].

Recent claims. On 8 September 2026 OpenAI announced a 166-page proof, produced by what it calls an internal AI system, that for every viscosity a smooth force confined in space and time can drive a fluid starting from rest to unbounded velocity in finite time with bounded kinetic energy; the paper presents this as establishing statements (C) and (D), and the same announcement claims blow-up for the unforced Euler equations from smooth data [10, 13]. OpenAI describes "an analytical proof and a Lean formalization", released the Lean code, and states: "We do not intend to claim the Millennium Prize for this result" [10]. On 11 September CMI wrote that it contemplates "the announcement that the Navier-Stokes problem has apparently been settled", in a process that "is deliberately unhurried" [11]. Zhen Lei and Xiao Ren "regard OpenAI's work as a major advance", and Córdoba and Martínez-Zoroa wrote that the result, "if verified under the stated hypotheses", would establish an allowed alternative but "is distinct from unforced Navier–Stokes blow-up" [12, 13]. Alexander Bastounis, Fabian Circelli and Anders Hansen argue in a preprint that "the formalised Lean proof does not correspond to the NL proof" (the natural-language proof), while Thomas Hales wrote that the formalization "has already been confirmed by more than a dozen proof-checkers" and that a human should still check its statement against Fefferman's [14, 15].

Main approaches

Regularity criteria (towards A and B). Show that some quantity that controls smoothness can never become infinite. Strongest evidence: partial regularity [2, 4] and the L³ criterion [3]. Main difficulty: Fefferman judged that "it appears to be very hard to go further" than partial regularity [4], and Tao's averaged equation shows that the energy identity and harmonic-analysis estimates alone cannot suffice [5].

Self-similar and computer-assisted blow-up. Find a profile that contracts in a fixed shape, then prove that a true solution stays close to it. Strongest evidence: the computer-assisted proof for Boussinesq and Euler with a boundary [7] and high-precision unstable profiles [8]. Main difficulty: the proofs for smooth data use a boundary, and constructions in all of space have needed data that are not fully smooth [7, 13]; without a boundary, the singularities sought are expected to be unstable, requiring initial conditions "tuned with infinite precision" [8].

Cascades of amplification. Build the flow in stages, each adding a more concentrated layer that the previous one amplifies, while keeping the force smooth [9, 13]. Strongest evidence: forced blow-up for the porous media, Boussinesq and Euler equations, and for weakly dissipative Navier–Stokes [9, 13]. Main difficulty: dissipation damps each new layer; the authors write that "the ordinary Laplacian cannot be reached by simply rescaling this construction" [13].

Forced blow-up with oscillatory pulses. OpenAI's construction drives a contracting, increasingly slender swirling core and adds oscillatory pulses whose momentum fluxes cancel the singular part of the force the core would otherwise need, so that the remaining force stays smooth [10]. Strongest evidence: the announced proof and its formalization, examined in the responses above [10, 12, 14, 15]. Main difficulty: it relies on an external force, so it concerns statements (C) and (D), not the unforced (A) and (B) [10, 13].

Open sub-questions

  1. Without a force. Can smooth, finite-energy initial data blow up with no force at all, settling (A) or (B) negatively, for instance along Tao's programme or by adapting the forced constructions or the claimed unforced Euler blow-up [5, 9, 10]?
  2. Stronger dissipation. For which fractional dissipation orders between about 0.093 and 2 can forced blow-up be constructed [13]?
  3. From unstable profiles to proofs. Can the high-precision unstable self-similar solutions for Euler with boundary be turned into computer-assisted proofs, and do unstable profiles exist without a boundary [8]?
  4. Beyond partial regularity. Can the Caffarelli–Kohn–Nirenberg bound on the singular set of suitable weak solutions be improved [2, 4]?
  5. Uniqueness. How far can non-uniqueness of Leray solutions be pushed from the borderline of the well-posedness theory towards smooth forces and data [6]?
  6. Checking a formal statement. Does the Lean statement released with the 2026 forced construction match each hypothesis of Fefferman's statement (C), including smoothness, decay and bounded energy [4, 14, 15]?

Work on this in Papex

If you have a testable answer to one of the sub-questions above, you can draft it as a private hypothesis in Papex, or prepare the draft from Claude or ChatGPT with the Papex connector. You review the exact version, sign it and decide whether to publish it; a public record that names this problem is listed on this page. A Papex record shows who signed which version and when. It does not check a proof, and it is not a submission for the Clay Mathematics Institute prize.

Key references

  1. Leray, J. (1934). Sur le mouvement d'un liquide visqueux emplissant l'espace. Acta Mathematica 63, 193–248. 10.1007/BF02547354 · Checked 14 Oct 2026
  2. Caffarelli, L., Kohn, R. & Nirenberg, L. (1982). Partial regularity of suitable weak solutions of the Navier–Stokes equations. Communications on Pure and Applied Mathematics 35(6), 771–831. 10.1002/cpa.3160350604 · Checked 14 Oct 2026
  3. Escauriaza, L., Seregin, G. A. & Šverák, V. (2003). L3,∞-solutions of the Navier–Stokes equations and backward uniqueness. Russian Mathematical Surveys 58(2), 211–250. 10.1070/RM2003v058n02ABEH000609 · Checked 14 Oct 2026
  4. Fefferman, C. L. Existence and smoothness of the Navier–Stokes equation. Official problem description, Clay Mathematics Institute (hosted PDF created 2006). Source · Checked 14 Oct 2026
  5. Tao, T. (2016). Finite time blowup for an averaged three-dimensional Navier–Stokes equation. Journal of the American Mathematical Society 29(3), 601–674 (published online 2015). 10.1090/jams/838 · Checked 14 Oct 2026
  6. Albritton, D., Brué, E. & Colombo, M. (2022). Non-uniqueness of Leray solutions of the forced Navier–Stokes equations. Annals of Mathematics 196(1), 415–455. 10.4007/annals.2022.196.1.3 · Checked 14 Oct 2026
  7. Chen, J. & Hou, T. Y. (2025). Singularity formation in 3D Euler equations with smooth initial data and boundary. Proceedings of the National Academy of Sciences 122(27), e2500940122. 10.1073/pnas.2500940122 · Checked 14 Oct 2026
  8. Wang, Y., Bennani, M., Martens, J., Racanière, S., Blackwell, S., Matthews, A. et al. (2025). Discovery of unstable singularities. Preprint, arXiv:2509.14185 (version 1, 17 September 2025), not peer reviewed. Source · Checked 14 Oct 2026
  9. Tao, T. (2026). Finite time blowup with smooth forcing term for the incompressible porous medium, Boussinesq, and incompressible Euler equations. Blog post, What's new, 7 September 2026. Source · Checked 14 Oct 2026
  10. OpenAI (2026). Finite time blowup for Navier–Stokes. Preprint PDF, 166 pages, 8 September 2026, Source · Checked 14 Oct 2026
  11. Clay Mathematics Institute. Navier–Stokes Equation (problem page), Navier-Stokes Announcement (news, 11 September 2026) and Rules for the Millennium Prize Problems. Source · Checked 14 Oct 2026
  12. Lei, Z. & Ren, X. (2026). Finite-time blowup for Navier–Stokes with smooth forcing, Part I: construction of self-similar solutions with admissible stress and flat remainder. Expository preprint, arXiv:2609.35406 (version 2, 29 September 2026), not peer reviewed. Source · Checked 14 Oct 2026
  13. Córdoba, D. & Martínez-Zoroa, L. (2026). On classical solutions and singularity formation in incompressible fluids. Guest post on T. Tao's blog What's new, 4 October 2026. Source · Checked 14 Oct 2026
  14. Bastounis, A., Circelli, F. & Hansen, A. C. (2026). Navier–Stokes lost in translation: why Lean verification of AI autoformalisation does not guarantee correct natural language proofs. Preprint, arXiv:2610.08144 (version 1, 6 October 2026), not peer reviewed. Source · Checked 14 Oct 2026
  15. Hales, T. (2026). What mathematicians should know about the Lean Theorem Prover: questions of reliability and AI. Guest post on T. Tao's blog What's new, 9 October 2026. Source · Checked 14 Oct 2026

Page reviewed 14 Oct 2026

Source

Under review, as listed by Clay Mathematics Institute · Millennium Prize Problems. Checked 14 Oct 2026.

The status is the source’s. Papex does not decide whether a problem is solved.

Status history

  1. Under review · Clay Mathematics Institute · Millennium Prize Problems · Checked 14 Oct 2026
  2. Open · Clay Mathematics Institute · Millennium Prize Problems · Checked 29 Sep 2026

Prize

Clay Mathematics Institute · $1,000,000

Prize rules

Only Clay Mathematics Institute awards this prize, under its own rules. Your Papex record is separate: a signed, dated record of your work, not a prize entry.

Records on this problem

No public record addresses this problem yet.

More in Mathematics