Fernandez v. United States, 608 U.S. ___ (2026)

NOTICE: This opinion is subject to formal revision before publication in the United States Reports. Readers are requested to notify the Reporter of Decisions, Supreme Court of the United States, Washington, D. C. 20543, [email protected], of any typographical or other formal errors.

SUPREME COURT OF THE UNITED STATES

_________________

No. 24–556

_________________

JOE FERNANDEZ, PETITIONER v. UNITED STATES

on writ of certiorari to the united states court of appeals for the second circuit

[May 28, 2026]

Justice Barrett delivered the opinion of the Court.

A federal statute authorizes district courts to shorten prison sentences for “extraordinary and compelling reasons”—most commonly, conditions like the prisoner’s age and infirmity. 18 U. S. C. §3582(c)(1)(A). Petitioner Joe Fernandez argues that doubts about a conviction’s validity also qualify as “extraordinary and compelling” reasons for relief under the statute. We disagree. A prisoner who collaterally attacks the validity of his conviction must proceed through 28 U. S. C. §2255, not 18 U. S. C. §3582.

I

Fernandez was indicted in 2013 for his role in the assassination of two gang members. The prosecution’s theory was that members of a drug ring had paid Fernandez $40,000 to act as the backup shooter. When the primary shooter’s gun jammed, Fernandez stepped in. He fired 14 rounds, killing both victims.

Fernandez’s cousin and alleged co-conspirator, Patrick Darge, testified against him at trial. To discredit Darge, Fernandez’s counsel argued that Darge was framing Fernandez to protect the real second shooter, Darge’s brother. Unconvinced, the jury convicted Fernandez of murder for hire and a firearms offense. See 18 U. S. C. §§924(j), 1958. The District Court then imposed two consecutive life sentences.

Fernandez asked the District Court to reconsider, alleging that the prosecution had violated Brady v. Maryland, 373 U.S. 83 (1963). He pressed that argument several times, first in a motion to vacate the verdict, then in a motion for new trial, and again in a motion for rehearing. According to Fernandez, the Government should have disclosed that another alleged co-conspirator, Luis Rivera, had denied driving the getaway car. To determine whether Rivera had in fact denied his involvement, the District Court reviewed the Government’s notes from Rivera’s interview; it concluded that they “d[id] not contain any relevant information.” United States v. Fernandez, 2014 WL 7180225, *1 (SDNY, Nov. 25, 2014). There was no evidence that Rivera had ever made such a statement, and it was not clear that it would have mattered to the jury if he had. Even so, the judge noted that he was “troubled” that prosecutors had “allow[ed] Rivera to plead to a lesser narcotics offense and [had] dismiss[ed] the murder and gun charges against him.” Id., at *3. In the judge’s view, the prosecutors could have used the evidence in Fernandez’s case to show that Rivera had engaged in similar wrongdoing. Ibid.

Fernandez appealed, and the Second Circuit affirmed his conviction and sentence. See United States v. Fernandez, 648 Fed. Appx. 56 (2016). Although Fernandez argued that the Government had failed to turn over its notes before trial, those notes “d[id] not reflect Rivera’s unequivocal denial of a getaway driver role.” Id., at 61. “[I]n any event,” the court said, “Fernandez fails to show how the notes could have been ‘useful’ ” for impeaching any witness. Ibid. Fernandez also argued that insufficient evidence supported his conviction: He insisted that Darge’s testimony was not credible and highlighted the Government’s choice not to charge Rivera as a co-conspirator in the murder-for-hire plot, which he considered indicative of the weakness of the Government’s case. But the court rejected the insufficiency claim too. It held that a reasonable jury could credit Darge’s testimony, that it was corroborated by other evidence in the record, and that the evidence was sufficient to support Fernandez’s conviction.

Fernandez persisted. He twice moved for postconviction relief under 28 U. S. C. §2255. The first time, Fernandez argued (among other things) that he was actually innocent because the testimony against him was not credible. He said that “[e]ven the Government did not fully believe Patrick Darge,” as shown by the Government’s decision not to charge Rivera with murder for hire. Reply Brief in No. 18–06 (CA2), pp. 8–9, n. 2. The Second Circuit described this argument as “plainly meritless,” concluding that “[t]he jury was entitled to credit the witnesses who testified that Fernandez committed the crimes with which he was charged.” Fernandez v. United States, 757 Fed. Appx. 52, 55–56 (2018). The second collateral challenge succeeded. The District Court set aside Fernandez’s conviction for a firearms offense after United States v. Davis, 588 U.S. 445 (2019), held that 18 U. S. C. §924(c)(3)’s residual clause is unconstitutionally vague. See Fernandez v. United States, 569 F. Supp. 3d 169 (SDNY 2021).

Still, the District Judge remained troubled about Fernandez’s murder-for-hire conviction. No new evidence had surfaced. But the judge continued to focus on Rivera—the getaway driver who had pleaded guilty to a less serious charge. In his order vacating the firearms conviction, the judge speculated that the Government had offered Rivera such a lenient plea deal because it “kn[ew] something” inconsistent with Darge’s testimony. Id., at 174, n. 4. And if Darge was lying, the judge wondered, “perhaps there is something to Petitioner’s argument, that not he, but Patrick Darge’s brother . . . was the second shooter, and that Patrick Darge testified to cover that up.” Ibid. Notably, the Second Circuit had already rejected this argument twice, and Fernandez had not briefed it again before the District Court. But the judge brought it up anyway, pointedly noting that if Fernandez’s life sentence on the murder-for-hire charge “were to be commuted, or held unlawful, [Fernandez] would be released immediately.” Id., at 179.

Fernandez took the hint. Within two weeks, he filed a motion for compassionate release under §3582(c)(1)(A)(i). See No. 1:10–cr–00863 (SDNY), ECF Doc. 248. He argued that extraordinary and compelling reasons—above all, that he was innocent—warranted a sentencing reduction. According to Fernandez, Darge’s testimony was biased and uncorroborated by forensic evidence; moreover, his identification of Rivera as the getaway driver was inconsistent with a later statement made by Rivera himself. The Government had not turned over its notes on that statement, Fernandez asserted, and had allowed Rivera to plead guilty to a lesser charge.

The District Court granted the motion. Nearly a decade after presiding over the trial and sentencing, the judge felt “a certain disquiet” about whether Darge’s testimony had been truthful and whether the cross-examination had been effective. 2022 WL 17039059, *4 (SDNY, Nov. 17, 2022). He had “strong concerns” about the Government’s decision to charge the getaway driver with a lesser offense and “doubt[ed] that the jury’s verdict was correct.” Ibid. Darge might have “sacrifice[d]” Fernandez “to save his brother,” and Fernandez might not have been the shooter or even “a member of the conspiracy.” Ibid. That “disquiet” formed the “basis of [the District Court’s] finding that Petitioner Joe Fernandez has shown extraordinary and compelling circumstances for his release.” Ibid.[2]

The Second Circuit reversed, holding that “challenges to the validity of a conviction are not cognizable as ‘extraordinary and compelling reasons’ under section 3582(c)(1)(A).” 104 F. 4th 420, 431 (2024). Congress has channeled such claims into federal postconviction proceedings under 28 U. S. C. §2255, and by design, compassionate release and §2255 do not overlap. “Challenging the validity of a conviction under the extraordinary-and-compelling-reasons prong of section 3582,” the court explained, “would permit a defendant to ‘evade the collateral review structure’ of section 2255.” 104 F. 4th, at 430 (alteration omitted).

While seven other Circuits have reached the same conclusion, two have taken the other side.[3] We granted certiorari to resolve the split. 605 U.S. 931 (2025).

II

Collateral attacks on federal convictions are governed by 28 U. S. C. §2255, which imposes tight procedural constraints. It carries a one-year statute of limitations. §2255(f ). With very narrow exceptions, a prisoner gets only one shot at collateral relief. §§2255(h)(1)–(2). Claims that have already been raised and rejected on direct review typically cannot be relitigated in §2255 motions. See Kaufman v. United States, 394 U.S. 217, 227, n. 8 (1969); Reed v. Farley, 512 U.S. 339, 358 (1994) (Scalia, J., concurring in part and concurring in judgment). And claims not raised on direct review are procedurally defaulted unless the prisoner can demonstrate “ ‘cause’ ” and “ ‘prejudice,’ ” or else actual innocence. Bousley v. United States, 523 U.S. 614, 622 (1998). If the prisoner clears these procedural hurdles, he must demonstrate that the applicable substantive law entitles him to relief. Section 2255’s companion statute, 28 U. S. C. §2254, imposes similar constraints on state prisoners.

Recall that Fernandez brought—and lost—a challenge to his conviction under §2255. This case arises from his motion presenting similar arguments under a different statute, 18 U. S. C. §3582, which permits prisoners to seek compassionate release from prison. In contrast to §2255’s reticulated scheme, §3582 imposes a single procedural requirement: A prisoner must first present his request for release to the Bureau of Prisons. §3582(c)(1)(A). A motion may then be filed in district court, where the prisoner (or the Bureau, acting on his behalf ) must show that “extraordinary and compelling reasons” warrant his early release from confinement. §3582(c)(1)(A)(i).

This is not the first time we have addressed the relationship between the habeas statutes and other kinds of prisoner litigation. In Preiser v. Rodriguez, prisoners brought claims under 42 U. S. C. §1983 challenging the forfeiture of their good-time credits, and their claims “plainly came within the literal terms of that statute.” 411 U.S. 475, 488 (1973). Even so, we held that “[t]he broad language of §1983” did not afford the prisoners a cause of action. Id., at 489. “Congress has determined that habeas corpus is the appropriate remedy for state prisoners attacking the validity of the fact or length of their confinement, and that specific determination must override the general terms of §1983.” Id., at 490. Challenges “close to the core of habeas corpus” must be brought, if at all, under “the specific federal habeas corpus statute”—in that case §2254—which was “explicitly and historically designed to provide the means for a state prisoner to attack the validity of his confinement.” Id., at 489. A contrary result would allow prisoners to evade §2254’s requirements, and so would “wholly frustrate explicit congressional intent.” Ibid. That is why “§1983 must be read in harmony with the habeas statute.” District Attorney’s Office for Third Judicial Dist. v. Osborne, 557 U.S. 52, 66 (2009).

Gonzalez v. Crosby, 545 U.S. 524 (2005), flows in the same vein. There, we held that prisoners may not use Federal Rule of Civil Procedure 60(b) to argue that a denial of habeas relief was wrong on the merits and warrants reopening. The language of Rule 60(b) is expansive, permitting courts to reopen a final judgment whenever a movant shows “any . . . reason that justifies relief.” Fed. Rule Civ. Proc. 60(b)(6) (emphasis added). And the text of §2254 “did not expressly circumscribe the operation of Rule 60(b).” Gonzalez, 545 U. S., at 529. Yet allowing movants to use a Rule 60(b) motion to make merits arguments would still collide with §2254. It would permit movants to “circumvent” the strict statutory habeas standards, including “that a successive habeas petition be precertified by the court of appeals.” Id., at 531–532 (citing 28 U. S. C. §2244(b)(3)). And the Rule could not be applied to habeas proceedings in a manner “ ‘inconsistent with’ ” §2254. 545 U. S., at 529 (quoting 28 U. S. C. §2254 Rule 11 (2000 ed., Supp. IV)). Thus, Rule 60(b)’s allowance of reopening for “any . . . reason” was narrowed to account for §2254.

Much like the prisoners in Preiser and Gonzalez, Fernandez has pursued a collateral attack outside the rigorous habeas framework. And as in those cases, the workaround is impermissible. Challenging the validity of a conviction through a compassionate release motion circumvents the exacting requirements of §2255.[4]

Indeed, Fernandez candidly admits that avoiding §2255’s procedural and substantive requirements is the benefit of his approach. As for procedure: Fernandez asserted at oral argument that “there are harsh limitations associated with [§]2255, and Congress has stated . . . that those limitations are appropriate, but it doesn’t mean that in every case, as applied to every defendant, that those limitations are fair.” Tr. of Oral Arg. 18–19. According to Fernandez, the ability to bring a merits-based challenge under §3582 mitigates “[t]he error and the unfairness of [the prisoner’s] not having been able to bring it under [§]2255.” Id., at 19. His take on substance is similar. He offers the example of a prisoner who claims that prosecutors withheld mitigating evidence and that his lawyer was constitutionally ineffective for failing to pursue that evidence. Fernandez posits that these arguments would fail to satisfy “the legal intricacies of the doctrines related to Brady v. Maryland and Strickland v. Washington.” Brief for Petitioner 41 (citation omitted). Still, Fernandez asserts, these half-baked legal challenges could count as “extraordinary and compelling reasons” that justify a reduced sentence. Ibid. The dissent, too, proclaims that the “core mission” of compassionate release is to offer deserving prisoners relief from the rigors of §2255. Post, at 21 (opinion of Jackson, J.).

It is easy to see why Fernandez would prefer this construction. Section 3582—in sharp contrast to §2255—imposes no time limit or numerical cap on motions. Nor is there any bar to raising issues waived or already decided on direct appeal. So a prisoner proceeding under §3582 could challenge the validity of his conviction repeatedly, for years after it became final, even if the issue had already been raised or decided in prior proceedings. This system would enable prisoners not only to bypass §2255, but also to override it. Consider this very case: Fernandez lost an argument on direct review and under §2255, but then repackaged it in a motion under §3582(c)(1)(A)(i) and obtained release from prison on those twice-rejected grounds.

We will not set §3582 and §2255 at cross-purposes when we can construe them “in harmony.” District Attorney’s Office, 557 U. S., at 66. Congress made compassionate release available only when “extraordinary and compelling reasons” justify it. §3582(c)(1)(A)(i). That is a demanding standard. And as we did with §1983 in Preiser and Rule 60(b) in Gonzalez, we interpret §3582 against the backdrop of the detailed process that Congress “explicitly and historically designed” for “attack[ing] the validity of [a prisoner’s] confinement.” Preiser, 411 U. S., at 489; see also Jones v. Hendrix, 599 U.S. 465, 473 (2023). Seen in that light, the supposed invalidity of a conviction is not among the “extraordinary and compelling reasons” that justify compassionate release.

While Congress has not defined the “extraordinary and compelling reasons” that may warrant a reduced sentence, these criteria are not empty vessels. “Extraordinary” means “most unusual,” “far from common,” and “having little or no precedent.” Webster’s Third New International Dictionary 807 (1976); see 5 Oxford English Dictionary 614 (2d ed. 1989) (“[o]ut of the usual or regular course or order”). “Compelling” means “tending to convince or convert by or as if by forcefulness of evidence.” Webster’s Third New International Dictionary, at 463; see 3 Oxford English Dictionary, at 600 (“irresistible; demanding attention, respect”). We will focus on the latter requirement.

An argument that is compelling in one context is not necessarily so in another; the force of an argument depends on what it seeks to justify. For instance, a 25th wedding anniversary is a convincing reason to shorten a business trip, but it is not a convincing reason to shorten a prison sentence. So here, we do not ask in the abstract whether a challenge to the validity of a conviction is “compelling.” We ask whether it is an especially convincing reason to grant the relief sought: compassionate release from prison.[5]

Section 2255’s reticulated scheme bears on this inquiry. A reason is not “compelling” if Congress has channeled it through the postconviction statutes. Even Fernandez recognizes that an argument may be less compelling if it could have been—and was not—asserted through §2255. He admits that courts may consider “as part of the extraordinary and compelling analysis” whether a “defendant sat on his rights or whether he’s taken all opportunities to enforce his rights.” Tr. of Oral Arg. 15–16. Yet the same is true for §2255’s other substantive and procedural requirements: That Congress has chosen to make §2255 a high bar is not a compelling reason to reduce the sentences of prisoners who fall short.

Other features of the statute point in the same direction. The name for §3582(c)(1)(A)—“Compassionate Release”—highlights its focus on granting mercy rather than righting legal wrongs. First Step Act of 2018, §603(b), 132Stat. 5239 (heading). In keeping with the theme of “compassion,” Congress in §3582 gives special protection to defendants with a “terminal illness.” §3582(d). For instance, the Director of the Bureau of Prisons must notify terminally ill prisoners of their right to seek compassionate release, §§3582(d)(2)(A)(i), (B)(i); must sometimes assist in preparing those compassionate release requests, §§3582(d)(2)(A)(iii), (B)(iii); and must report to Congress “the number of prisoners who died while their request was pending,” §3582(d)(3)(H). In addition, the statute expressly identifies one situation in which a sentencing reduction is warranted: when “the defendant is at least 70 years of age, has served at least 30 years” of a certain type of mandatory life sentence, and the Director of the Bureau of Prisons determines “that the defendant is not a danger to the safety of any other person or the community.” §3582(c)(1)(A)(ii). And in another statute that references §3582(c)(1)(A), Congress has permitted the Sentencing Commission to treat a defendant’s rehabilitation as a relevant consideration in granting compassionate release. 28 U. S. C. §994(t).

The role of the Bureau of Prisons reflects the statute’s focus on a defendant’s personal circumstances—his advanced age, safety risk, illness, rehabilitation, and the like. The Bureau evaluates compassionate release motions in the first instance. Until 2018, a district court could reduce a sentence only “ ‘upon motion of the Director of the Bureau of Prisons.’ ” Sentencing Reform Act of 1984, 98Stat. 1998–1999. After the First Step Act, prisoners may file their own compassionate release motions, but only if the Bureau declines or fails to respond to a prisoner’s request that the Director file on his behalf. 132Stat. 5239. Fernandez agrees that this procedural shift in the Bureau’s role—from exclusive filer to initial evaluator—did not change the substantive standard applicable to compassionate release motions. See Tr. of Oral Arg. 9–12.

If compassionate release were a vehicle for attacking the validity of a conviction, it is hard to see why the Bureau of Prisons would run the process. The Bureau’s institutional expertise lies in the daily lives of prisoners: matters relating to their “safekeeping,” “care,” “subsistence,” “protection,” “instruction,” and “discipline.” 18 U. S. C. §4042(a). So it makes sense for the statute to instruct the Bureau to help terminally ill prisoners seek compassionate release. §3582(d). By contrast, it makes little sense to have the Bureau evaluate legal arguments and comb trial records to determine whether a prisoner has a “compelling” argument that his conviction was wrongful. Even Fernandez concedes that analyzing potential trial errors is outside the Bureau’s bailiwick. Tr. of Oral Arg. 10. Take Fernandez’s own situation: How would the Bureau determine whether the testimony at his trial was credible? Or take Fernandez’s hypothetical: How would the Bureau evaluate arguments raised under Brady v. Maryland, 373 U.S. 83, and Strickland v. Washington, 466 U.S. 668 (1984)? Supra, at 8. Congress’s decision to route compassionate release motions through the Bureau strongly suggests that the “extraordinary and compelling reasons warrant[ing]” compassionate release must be reasons that the Director is competent to assess. §3582(c)(1)(A)(i).

Regulatory history points the same way. Congress has authorized the Sentencing Commission to issue policy statements about when compassionate release is warranted, 28 U. S. C. §994(t), and these policy statements shape whether courts may grant compassionate release motions, 18 U. S. C. §3582(c)(1)(A). For decades, the Sentencing Commission has tied the availability of compassionate release to a defendant’s personal circumstances. See United States Sentencing Commission, Guidelines Manual App. C, Amdt. 698 (Nov. 2007) (USSG) (listing “ ‘extraordinary and compelling reasons’ ” to include “ ‘terminal illness,’ ” “ ‘a permanent physical or medical condition,’ ” “ ‘[t]he death or incapacitation of the defendant’s only family members capable of caring for the defendant’s minor child,’ ” or “ ‘other’ ” reasons as determined by the Director of the Bureau of Prisons). While the Sentencing Commission has issued more detailed policy statements over the years, it has maintained the same basic categories.[6] See USSG §1B1.13(1), comment., n. 1 (Nov. 2021). It has never said that the invalidity of a conviction qualifies as an “extraordinary and compelling reaso[n] warrant[ing]” compassionate release. §3582(c)(1)(A)(i).

In short, the structural inference of Preiser and Gonzalez is borne out in §3582 itself. The heartland “extraordinary and compelling reasons” that might warrant an early release from prison—age, illness, a child left with no guardian—bear no resemblance to the grounds for relief under §2255. Fernandez cannot deploy §3582 to blunt what he perceives to be the sharp edges of §2255.

III

Fernandez insists that because §2255 and §3582 offer different forms of relief, the structural inference from Preiser and Gonzalez does not apply. Under §2255, a court could set aside Fernandez’s conviction as if it never happened. But Fernandez is not going for that—at least not in this round. Now, he seeks the more modest relief permitted by §3582: a reduced sentence. If he succeeds, he would leave prison early, but his murder-for-hire conviction—along with its collateral effects—would remain in place. That distinction, he says, means that he is not circumventing §2255, but rather proceeding on an entirely different track. See also post, at 11 (Jackson, J., dissenting).

Instead of helping him, this argument highlights the mismatch between the error Fernandez alleges and the remedy he seeks. When an elderly or infirm prisoner is discharged early, the reasons for release have nothing to do with the lawfulness of the conviction—so it makes sense to leave it in place. But when a prisoner persuades a court that his conviction is invalid, the remedy of a little less prison time does not redress the wrong. If a conviction is invalid, the fitting remedy is to vacate it, as §2255 allows. That §3582 does not offer this remedy is evidence that it is not a suitable vehicle for the claim.

In any event, the difference in relief does not render Preiser irrelevant. The remedies were not identical in Preiser either: The prisoner sought an injunction under §1983 for the restoration of good-time credits, rather than habeas relief. This formal distinction did not change our analysis; what mattered was that Preiser brought a claim “close to the core of habeas corpus.” 411 U. S., at 489. When, as here, a prisoner “attack[s] the validity of his confinement” and “seeks either immediate release from that confinement or the shortening of its duration,” his claim belongs under the umbrella of the postconviction statutes. Ibid.

And to be clear: Fernandez is challenging the validity of his conviction, even though he is not asking to have it vacated or set aside. Whether a prisoner’s request for relief challenges a conviction turns on the nature of the arguments presented. See, e.g., Gonzalez, 545 U. S., at 531, 533 (holding that only some arguments presented in a Rule 60(b) motion would improperly infringe on §2254). When a prisoner moves for compassionate release based on a terminal illness, his motion does not suggest that his conviction is invalid. But if a prisoner moves for relief because of alleged problems with the underlying criminal investigation or trial, his motion suggests that his conviction is flawed. That is true here: Fernandez’s motion said he “would never commit” murder for hire; questioned whether the District Judge would have reached the same verdict as the jury; urged the court not to credit Darge’s testimony; alleged that the Government failed to turn over or present to the jury “highly exculpatory” notes from interviewing Rivera; and even argued that Fernandez could sufficiently show his actual innocence to satisfy Schlup v. Delo, 513 U.S. 298 (1995), which governs certain procedurally defaulted habeas claims. See ECF Doc. 248, pp. 7–9, 14. The District Court then impermissibly treated Fernandez’s challenge to his conviction as an “extraordinary and compelling reaso[n]” for sentencing reduction. 18 U. S. C. §3582(c)(1)(A)(i). It expressed “[r]eason to [q]uestion the [v]erdict,” “doubt that the jury’s verdict was correct,” and the belief that the prosecution distrusted the testimony of its star witness. 2022 WL 17039059, *4. In other words, the District Court harbored doubts about the soundness of the conviction.[7]

Fernandez offers a backup position: Even if §3582(c)(1)(A)(i) does not ordinarily allow district courts to grant compassionate release for reasons that would support a §2255 motion, he asserts that actual innocence claims are an exception. He points out that “a factual showing of actual innocence ‘ha[s] never been held to state a ground for federal habeas relief absent an independent constitutional violation.’ ” Brief for Petitioner 41–42 (quoting Herrera v. Collins, 506 U.S. 390, 400 (1993)). Thus, he says, such a claim must be cognizable under §3582; otherwise, a prisoner might have nowhere to go. See also post, at 9–10, 17–18 (Jackson, J., dissenting).

It bears emphasis that this argument is only marginally relevant to Fernandez himself. No court—not even the District Court that reduced his sentence—concluded that Fernandez was actually innocent. His evidentiary challenges were repeatedly rejected, and the District Court reduced his sentence because of “disquiet” about the conviction, not confidence in Fernandez’s innocence. 2022 WL 17039059, *4. So even if an actual innocence claim were cognizable under §3582, Fernandez would not benefit from it.

In any event, this case gives us no occasion to decide the antecedent question whether a prisoner may assert a freestanding actual innocence claim under §2255. We have never ruled that possibility out and do not do so now. See Herrera, 506 U. S., at 417. Instead, we hold simply that a compassionate release motion is not a vehicle for raising such a claim. A prisoner asserting actual innocence “challenge[s] the validity of his conviction,” id., at 406, whether factually or legally, and that lies “close to the core of habeas corpus,” Preiser, 411 U. S., at 489. If it is “not in substance a ‘habeas corpus application,’ ” it is “at least similar enough that failing to subject it to the same requirements would” forge a path for bypassing the postconviction statutes. Gonzalez, 545 U. S., at 531. And for the reasons we have already stated, §3582 does not provide that kind of shortcut.

Fernandez’s remaining arguments echo those presented in another case, which we also decide today. See Rutherford v. United States, ___ U. S. ___ (2026). Petitioners in both cases contend that the phrase “extraordinary and compelling reasons” vests courts with broad discretion to grant a prisoner compassionate release for virtually any reason whatsoever. We reject these arguments here for the same reasons we reject them in Rutherford. Id., at ___–___ (slip op., at 11–15). “While the terms ‘extraordinary’ and ‘compelling’ leave room for judgment, they are not so flexible as to encompass any consideration.” Id., at ___ (slip op., at 11).

*  *  *

The compassionate release provision is not a vehicle for attacking the validity of a conviction. Accordingly, the judgment of the Court of Appeals for the Second Circuit is affirmed.

It is so ordered.

SUPREME COURT OF THE UNITED STATES

_________________

No. 24–556

_________________

JOE FERNANDEZ, PETITIONER v. UNITED STATES

on writ of certiorari to the united states court of appeals for the second circuit

[May 28, 2026]

Justice Sotomayor, with whom Justice Kagan joins, concurring in the judgment.

The First Step Act of 2018 expanded district courts’ authority to “reduce [a] term of imprisonment” upon a finding that “extraordinary and compelling reasons warrant such a reduction” and that the reduced sentence is consistent with traditional sentencing factors. 18 U. S. C. §3582(c)(1)(A). For many of the reasons explained in Justice Jackson’s dissenting opinion, the Court’s decision interpreting that provision today is wrong. The majority improperly narrows the scope of §3582(c)(1)(A) based on tenuous inferences drawn from the federal habeas statutes. Its habeas-based rule distorts and expands the Court’s habeas-channeling cases, creating an atextual limitation found nowhere in §3582(c)(1)(A). In doing so, it superimposes a technical, and likely unworkable, habeas analysis on top of a sentence- reduction framework that broadly calls for holistic review. Moreover, the rule adopted today risks shutting out meritorious claims for a sentence reduction simply because they might resemble, in some amorphous way, a claim that might call into question a sentence or conviction.

Even so, the judgment below should be affirmed on a different, far simpler ground, which follows much more naturally from the statute before the Court: A motion for compassionate release cannot justify a reduced sentence if it relies solely on facts a court already considered in imposing the initial sentence, rather than any changed circumstances that developed after sentencing.

In general, a “court may not modify a term of imprisonment once it has been imposed.” §3582(c). Section 3582(c)(1)(A) authorizes a limited departure from that rule of finality when “extraordinary and compelling reasons” arise so as to “warrant . . . a reduction” of the sentence imposed. For a reason to “warrant” a change of a sentence, however, it cannot have been a consideration that was part of the initial sentencing determination to begin with. At the very least, changing a sentence already imposed under §3582(c)(1)(A) requires something to have changed, after sentencing, in a way that alters the sentencing calculus. See, e.g., Setser v. United States, 566 U.S. 231, 242–243 (2012) (explaining that §3582(c)(1)(A) provides for relief when “ ‘developments that take place after the first sentencing’ produc[e] unfairness to the defendant” (citation omitted)). Otherwise, district courts would have indefinite authority to revise sentences already imposed based on information and arguments that were previously considered and rejected. Such authority would seriously diminish the finality of sentences required by §3582(c).

The District Court’s grant of relief here ran afoul of this principle. Petitioner Joe Fernandez has been litigating the issues at the core of his §3582(c)(1)(A) motion since his trial. From the start, Fernandez has contended that the evidence against him was weak and that the prosecution failed to prove beyond a reasonable doubt that he (and not someone else) committed the murders for which he was convicted. He raised these issues at his trial, see ante, at 1–2, at his sentencing, see App. 57–62, in a motion to reconsider his sentence, see ante, at 2–3, and in a postconviction motion under 28 U. S. C. §2255, even though “[n]o new evidence had surfaced,” ante, at 3–4. Each effort failed in the trial court and on appeal. Yet these same arguments finally found success under §3582(c)(1)(A) based on the District Court’s renewed feelings of “ ‘doub[t]’ ” and “ ‘disquiet’ ” about the original trial evidence supporting Fernandez’s jury verdict. Ante, at 4.*[2]

This grant of relief was inappropriate given the absence of any postsentencing developments, and the Court needed to go no further to affirm the decision below. Unfortunately, bad facts often make bad law, and so they did today. Correctly perceiving a problem with the District Court’s grant of relief, the majority responds with a rule that goes far beyond both what is needed to resolve this case and, worse, what the text and relevant precedents can bear. Because I do not agree with the majority’s reasoning in support of its decision to affirm the decision below, I concur only in the judgment of the Court.

SUPREME COURT OF THE UNITED STATES

_________________

No. 24–556

_________________

JOE FERNANDEZ, PETITIONER v. UNITED STATES

on writ of certiorari to the united states court of appeals for the second circuit

[May 28, 2026]

Justice Jackson, dissenting.

Section 3582(c)(1)(A) of Title 18 of the U. S. Code—known as the compassionate-release statute—authorizes district courts to reduce federal prison sentences under specified circumstances. This case is about the scope of that authority. By its terms, the compassionate-release statute is flexible, permitting sentence reductions for “extraordinary and compelling reasons.” 18 U. S. C. §3582(c)(1)(A)(i). The breadth of this language is no accident: It reflects Congress’s intent to preserve some of the traditional discretion afforded to district courts to ensure just treatment of defendants in criminal cases.

Today, the Court arbitrarily restricts that discretion by grafting an atextual rule onto §3582(c)(1)(A). It holds that a district court may not base its “extraordinary and compelling” findings on any reason that “collaterally attacks the validity” of the prisoner’s conviction. Ante, at 1. Such “attacks,” the majority says, must be brought through motions for habeas relief under 28 U. S. C. §2255. Ibid. But this restriction comes of out nowhere—it finds no support in the statute’s text or history, nor can it be justified by our precedents.

What, then, explains the new categorical rule that the Court adopts today? The answer appears to be the majority’s intuition that the District Court’s grant of petitioner Joe Fernandez’s compassionate-release motion qualifies as an abuse of discretion under the circumstances presented here. That may well be true—but not because of an implicit, habeas-based limitation on the reach of §3582(c)(1)(A).

Because the Court of Appeals erroneously relied on such a habeas-based rule in reversing the District Court’s grant of compassionate release, and the majority now endorses that approach, I respectfully dissent. I would vacate, rather than affirm, the decision of the Court of Appeals.

I

The majority holds that the District Court erred in granting Fernandez’s compassionate-release motion because §3582(c)(1)(A) contains an implicit habeas-based limitation: Sentence reductions that derive from purported “[c]ollateral attacks on federal convictions” are not authorized. Ante, at 5. It is not clear whether the majority believes, on the one hand, that conviction-related concerns can never be sufficiently “extraordinary and compelling” or, on the other, that even if there are “extraordinary and compelling” reasons to doubt the defendant’s conviction, a sentence reduction is nevertheless unavailable.

What is evident is that, in devising this new rule, the majority devotes most of its analysis to describing the contours of habeas relief under §2255 and gives short shrift to the text and history of §3582(c)(1)(A). Had the majority focused instead on the statute actually before it, it would have seen that the compassionate-release provision operates as an independent stop-gap measure, designed to preserve a district court’s discretion to reduce a criminal sentence when “extraordinary and compelling” circumstances warrant that result. And when §3582(c)(1)(A) is thus properly understood, it is clear that the existence of habeas review has no bearing whatsoever on whether a defendant can seek or obtain compassionate release.

A

Start with the words Congress used in §3582(c)(1)(A)—they provide the most obvious strike against the majority’s newfound rule. Section 3582(c)(1)(A)’s text, while authorizing a sentence reduction, does not refer to §2255 or habeas review—in any way, shape, or form. Of course, if Congress had intended §2255 to limit the availability of the compassionate-release remedy, it could easily have said so. This Court has often taken the lack of a cross-reference to mean that two distinct statutory provisions bear no relation to one another. See, e.g., Pugin v. Garland, 599 U.S. 600, 608 (2023); Azar v. Allina Health Services, 587 U.S. 566, 576–577 (2019).

Not only is there no explicit basis for the majority’s habeas-based rule, there is also no implicit one. Section 3582(c)(1)(A) is a straightforward grant of discretion: It vests district courts with authority to reduce a prisoner’s sentence for “extraordinary and compelling reasons.” The terms “extraordinary” and “compelling” are commonly understood to relate to a narrow and specific set of circumstances defined by degree, not type. See Rutherford v. United States, ___ U. S. ___, ___ (2026) (Sotomayor, J., dissenting) (slip op., at 7–8). As the majority acknowledges, “ ‘[e]xtraordinary’ means ‘most unusual,’ ‘far from common,’ and ‘having little or no precedent,’ ” while “ ‘compelling’ means ‘tending to convince or convert by or as if by forcefulness of evidence.’ ” Ante, at 9–10 (quoting Webster’s Third New International Dictionary 463, 807 (1976)). Neither definition implies the exclusion of any kind of consideration from a district court’s sentence-reduction analysis. Rather, they suggest that any factor can be the basis for compassionate release, so long as it is both sufficiently unusual and sufficiently forceful. See Romag Fasteners, Inc. v. Fossil Group, Inc., 590 U.S. 212, 215 (2020) (observing that this Court does not “usually read into statutes words that aren’t there”).

The statutory context confirms the absence of any type-based restriction on the sentence-modification authority §3582(c)(1)(A) conveys, much less the majority’s preferred habeas limit. With full awareness of §2255, Congress placed only two textual limitations on what constitutes an “extraordinary and compelling reason” to modify a defendant’s sentence. First, the resulting sentence reduction must be “consistent” with the Sentencing Commission’s policy statements. 18 U. S. C. §3582(c)(1)(A). Second, “[r]ehabilitation of the defendant alone shall not be considered an extraordinary and compelling reason” for a sentence reduction. 28 U. S. C. §994(t). Far from connoting an additional habeas-based constraint, Congress’s express inclusion of these two restrictions on a district court’s “extraordinary and compelling” findings “ ‘implies the exclusion of othe[r]’ ” such limitations. Jennings v. Rodriguez, 583 U.S. 281, 300 (2018) (quoting A. Scalia & B. Garner, Reading Law 107 (2012)).

The majority offers little by way of a text-based comeback. Focusing on the word “compelling,” the majority observes that what “is compelling in one context is not necessarily so in another”; for example, a 25th wedding anniversary may be a “compelling” reason to shorten a business trip but not a prison sentence. Ante, at 10. While obviously true, that observation tells us nothing relevant. All agree that personal considerations (e.g., illness and family hardship) might warrant a sentence reduction under §3582(c)(1)(A), see ante, at 10–12, yet those considerations, too, are more or less “compelling” under various circumstances. Moreover, as explained in Part II–A, infra, Fernandez’s claim of actual innocence, if meritorious, would be a compelling reason to reduce his sentence. At the time the District Court considered his §3582(c)(1)(A) motion, Fernandez had already served 11 years in federal prison for a crime he says he did not commit. On the “extraordinary and compelling reasons” scale, his innocence claim is lightyears away from reducing a sentence to facilitate celebration of a wedding anniversary.

B

The text alone suffices to dispel the habeas-related limitation myth the majority adopts today. But, as is often true of statutes, it is not possible to fully understand the compassionate-release provision without appreciating its statutory and legislative history. See R. Katzmann, Judging Statutes 31 (2014) (“The task of the judge is to make sense of legislation in a way that is faithful to Congress’s purposes”). The historical context under which §3582(c)(1)(A) emerged and developed is “the water in which [the text] swim[s].” United States v. Hansen, 599 U.S. 762, 775 (2023). Notably absent from this history is any indication that Congress understood §2255 to implicitly limit the scope of compassionate release, either when it first enacted §3582(c)(1)(A) in 1984 or when it amended the provision in 2018.

1

Before 1984, “the Federal Government employed in criminal cases a system of indeterminate sentencing.” Mistretta v. United States, 488 U.S. 361, 363 (1989). Though “[s]tatutes specified the penalties for crimes,” they “nearly always gave the sentencing judge wide discretion to decide whether the offender should be incarcerated and for how long” or “whether restraint, such as probation, should be imposed instead of imprisonment or fine.” Ibid. The result: “great variation among sentences imposed by different judges upon similarly situated offenders” and “uncertainty as to the time the offender would spend in prison.” Id., at 366.

Congress set out to alleviate those problems with the Sentencing Reform Act of 1984. See 98Stat. 1987; see also S. Rep. No. 98–225, p. 65 (1983) (“The shameful disparity in criminal sentences is a major flaw in the existing criminal justice system, and makes it clear that the system is ripe for reform”). The Act’s most prominent reform was the creation of the U. S. Sentencing Commission, which Congress tasked with promulgating a new, determinate system of Sentencing Guidelines. See §217(a), 98Stat. 2017 (codified at 28 U. S. C. §991). Initially, the Guidelines were “binding on the courts.” See Mistretta, 488 U. S., at 367–368. Judges had some discretion to depart from an applicable Guidelines range, but only after “find[ing] an aggravating or mitigating factor present that the Commission did not adequately consider.” Id., at 367.[1]

By design, the new Guidelines-based system left little room for judges to exercise discretion when determining the applicable sentence. See S. Rep. No. 98–225, at 38 (describing the new system as aimed at eliminating “the unfettered discretion the law confers” on sentencing judges). At the same time, Congress recognized that removing all discretion would result in unfairly harsh sentences for some prisoners. Indeed, despite the uniformity aim of the Guidelines, lawmakers acknowledged the possibility of an “exceptional situation where someone obviously slips through the cracks and gets a much longer sentence.” Hearings on Revision of the Federal Criminal Code before the Subcommittee on Criminal Justice of the House Committee on the Judiciary, 96th Cong., 1st Sess., pt. 3, pp. 1842–1843 (1979) (Code Revision Hearings) (statement of Rep. Lungren). Federal judges similarly warned of “those occasional cases which cry out for some sort of revision, even though the sentence on that particular offender has been imposed by a perfectly conscientious sentencing court . . . and based upon facts and circumstances which were a matter of record.” See id., at 1902–1903 (statement of Hon. Harold R. Tyler).

This problem was what Congress set out to address when it created various “ ‘safety valve[s]’ ” on the back end of the sentencing process, allowing district courts to exercise their discretion to reduce sentences in “unusual case[s].” S. Rep. No. 98–225, at 121. The compassionate-release provision was one of them. As enacted in 1984, §3582(c)(1)(A) permitted the Director of the Bureau of Prisons (BOP) to petition a sentencing court for a reduction in a prisoner’s sentence. §212(a)(2), 98Stat. 1998–1999. And it authorized a court to grant such a petition if, “ ‘after considering the factors set forth in section 3553(a),’ ” the court found “ ‘that extraordinary and compelling reasons’ ” justified the reduction and that such a reduction was “ ‘consistent with applicable policy statements issued by the Sentencing Commission.’ ” Ibid.

Significantly for present purposes, Congress was well aware of §2255 in 1984. See, e.g., S. Rep. No. 98–225, at 234 (cross-referencing §2255 in explaining a different provision of the Sentencing Reform Act). But it never suggested that §3582(c)(1)(A) and §2255 were mutually exclusive, much less that habeas was the sole means of obtaining postsentencing relief from an unjust conviction. Nor did Congress indicate that the availability of compassionate release hinged on the unavailability of habeas review. Instead, it was clear from the start that compassionate release was a distinct vehicle for addressing potential inequities with respect to a defendant’s sentence: one specifically designed to allow for the flexible consideration of “extraordinary and compelling” circumstances notwithstanding the restrictions that otherwise made the sentencing system less discretionary. Id., at 121 (recognizing the “value” of “assur[ing] the availability of specific review and reduction of a term of imprisonment for ‘extraordinary and compelling reasons’ ”).

2

But, as it turned out, the compassionate-release safety valve rarely opened in the decades following §3582(c)(1)(A)’s enactment. Between 2006 and 2011, an average of “only 24 inmates” were “released each year through the BOP’s compassionate release program.” Dept. of Justice, Office of Inspector General, The Federal Bureau of Prisons’ Compassionate Release Program 1 (Apr. 2013). Over time, it became clear that BOP was drastically underutilizing its compassionate-release power, “resulting in inmates who may [have been] eligible candidates for release not being considered.” Id., at 11.

Of particular issue was the BOP’s exceedingly narrow interpretation of the statutory phrase “extraordinary and compelling,” which its Central Office staff had interpreted as referring only to “terminally ill medical conditions with a life expectancy of 1 year or less or those who are incapacitated and unable to perform their daily activities—criteria that [were] found nowhere in the relevant statute.” Id., at 13. And though BOP regulations technically “permit[ted] non-medical circumstances to be considered as a basis for compassionate release,” the agency “routinely reject[ed] such requests.” Id., at ii.

Congress took notice. In 2017, a bipartisan group of Senators wrote to the Department of Justice and expressed “dee[p] concer[n] that BOP [was] not fulfilling its role in the compassionate release process.” Letter from 12 U. S. Senators to J. Rod Rosenstein, Deputy Attorney General, & Dr. T. Kane, Acting BOP Director 3 (Aug. 3, 2017) (Letter from 12 U. S. Senators). Several of those Senators worked together later that year to introduce a “bipartisan sentencing and prison reform bill” known as the First Step Act, which passed the following year. 164 Cong. Rec. S7645 (Dec. 17, 2017) (statement of Sen. Durbin); see 132Stat. 5194. In a section entitled “Increasing the Use and Transparency of Compassionate Release,” the Act amended the compassionate-release provision to allow prisoners to file their own motions for sentence reductions, without approval from the BOP. §603(b), 132Stat. 5239.

This historical account can be summarized succinctly. Congress crafted §3582(c)(1)(A) to operate as a safety valve that would ameliorate the harshness of the new determinate scheme for sentencing. S. Rep. No. 98–225, at 121. It was not linked to, or limited by, the existence of habeas review in any respect. And that made perfect sense given the specific role Congress envisioned for compassionate release: allowing a discretionary sentence modification for a prisoner who presents “extraordinary and compelling” grounds for mercy.

II

A

The majority nevertheless insists on categorically excluding one potential ground for compassionate release—any consideration that supposedly “attacks the validity of [a] conviction” or sentence. Ante, at 1. One glaring flaw with this analysis is the majority’s failure to clearly establish when this forbidden circumstance arises; that is, what, exactly, does it mean for a prisoner to collaterally attack his conviction in the context of a compassionate-release motion? It is not hard to see how one man’s collateral attack is another man’s compelling bid for compassion.

Consider, for example, an actually innocent defendant—let’s call him Paul. Paul is serving a life sentence and, for reasons beyond his control, failed to discover indisputably exonerating evidence during his trial. But Paul cannot point to any way in which his conviction or sentence violates the Constitution or some other provision of the U. S. Code. He does not allege, for instance, that his counsel was ineffective or that the prosecution withheld exculpatory evidence. Habeas thus provides no avenue for relief. See 28 U. S. C. §2255(a) (authorizing relief for a defendant whose “sentence was imposed in violation of the Constitution or laws of the United States”).

An innocent man stuck in prison for life: Can there be a more “extraordinary and compelling” reason to shorten a prison sentence than that? And is there any logical reason to believe that Congress meant to prohibit Paul from petitioning the court for such relief under §3582(c)(1)(A)? Quite to the contrary, this manifest injustice would seem to be the quintessential situation in which the compassionate-release safety valve would apply. See Code Revision Hearings, at 1842–1843 (statement of Rep. Lungren) (expressing concern about federal prisoners “slip[ping] through the cracks”). But the majority’s rule would forbid it, simply and solely because §2255 exists.

Of course, this is not to say that every assertion of innocence will necessarily qualify as an “extraordinary” or “compelling” basis for a sentence reduction. For instance, a claim of innocence that simply rehashes the evidence presented at trial is unlikely to suffice. The concern that Fernandez is making such an unjustified innocence claim seems to be what drives much of the majority’s analysis today. See ante, at 16 (“No court—not even the District Court that reduced his sentence—concluded that Fernandez was actually innocent. His evidentiary challenges were repeatedly rejected”); see also ante, at 2 (Sotomayor, J., concurring in judgment) (“Petitioner Joe Fernandez has been litigating the issues at the core of his §3582(c)(1)(A) motion since his trial”). But the Court overcorrects for this concern and, in the process, adopts an atextual rule that categorically forbids any consideration of potential innocence whatsoever—no matter how extraordinary or compelling—either on its own or as part of a mix of other considerations. That result certainly does not follow from the text.

Nor would allowing consideration of potential innocence permit federal prisoners to circumvent §2255’s limitations on habeas relief, as the majority claims. See ante, at 7–9. That is because the habeas and compassionate-release provisions operate differently. To be sure, both statutes authorize the release of federal prisoners. But, where the statutory prerequisites are met, relief under §3582(c)(1)(A) is discretionary, whereas relief under §2255 is mandatory: The former provides that “the court . . . may reduce the term of imprisonment,” while the latter states that “the court shall vacate and set the judgment aside.” (Emphasis added.) Section 3582(c)(1)(A) thus provides no guaranteed end-run around §2255’s limitations on relief.

Section 3582(c)(1)(A) and §2255 also differ in the nature of the relief that they provide. Most notably, the grant of compassionate release—even on actual-innocence grounds—does nothing to invalidate the underlying conviction. A neighboring subsection explicitly clarifies that compassionate release does not disturb the finality of the “judgment of conviction.” §3582(b). Not so with habeas—when a §2255 applicant successfully challenges the validity of his conviction, the result is vacatur. See §2255(b) (“If the court finds . . . that there has been such a denial or infringement of the constitutional rights of the prisoner as to render the judgment vulnerable to collateral attack, the court shall vacate and set the judgment aside” (emphasis added)).

Put differently, habeas nullifies a conviction as a matter of law, while compassionate release shortens a sentence as a matter of grace. The majority does not adequately explain why these two forms of relief—which operate differently, exist for different reasons, and ultimately yield different results—must necessarily work to the exclusion of each other.

B

Despite these crucial observations, the majority forges ahead, insisting on grafting a habeas-based restriction onto the compassionate-release remedy. It gets there in two ways: first, by overstating the significance of the BOP’s role in order to bolster the conclusion that “personal circumstances” are the only cognizable factors under §3582(c)(1)(A); and, second, by overreading two of our prior habeas-channeling cases—Preiser v. Rodriguez, 411 U.S. 475 (1973), and Gonzalez v. Crosby, 545 U.S. 524 (2005). Neither move is availing.

1

Highlighting the BOP’s role in compassionate release and the fact that prison officials can observe things like “advanced age,” “safety risk,” and “illness,” the majority contends that Congress’s centering of the BOP “reflects the statute’s focus on a defendant’s personal circumstances.” Ante, at 11. But the majority cannot deny that Congress amended §3582(c)(1)(A) due to its concern that the BOP was mismanaging compassionate-release motions. See supra, at 8–9. The majority’s reliance on the BOP’s “institutional expertise,” ante, at 12, to interpret a statutory amendment aimed at reducing the BOP’s involvement is, at best, counterintuitive.

Recall that Congress amended the compassionate-release provision as part of the First Step Act—legislation that was broadly aimed at providing district courts with discretion to alleviate what many had come to believe were unduly harsh criminal sentences. See Concepcion v. United States, 597 U.S. 481, 486 (2022). Pointing back to the original, pre-First Step Act scope of §3582(c)(1)(A) does not account for Congress’s subsequent conclusion that BOP was applying the compassionate-release provision too narrowly. And when it invited district courts to consider §3582(c)(1)(A) motions filed directly by defendants, Congress did not say, or even suggest, that a sentence reduction under this provision is only available based on the kinds of factors BOP might raise.

Fernandez’s concession that the First Step Act “did not change the substantive standard applicable to” §3582(c)(1)(A), ante, at 11–12 (citing Tr. of Oral Arg. 9–12), is not to the contrary. That observation merely states the obvious: that the meaning of “extraordinary and compelling reasons” has always been expansive, leading Congress to conclude that the BOP had been misinterpreting “extraordinary and compelling” all along. See Letter from 12 U. S. Senators 3 (expressing “dee[p] concer[n]” about the BOP’s underutilization of compassionate release); §603(b), 132Stat. 5239 (titling the First Step Act’s amendment to §3582(c)(1)(A) “Increasing the Use and Transparency of Compassionate Release”). Fernandez seeks to redeem the statute’s meaning, not to change it.

The majority also tells only half the story when it declares that, “[f]or decades, the Sentencing Commission has tied the availability of compassionate release to a defendant’s personal circumstances.” Ante, at 12. To be sure, the agency has always identified certain personal characteristics (such as “ ‘a serious physical or medical condition’ ”). But it has also included a catchall category for any circumstances determined to be “ ‘extraordinary and compelling.’ ” United States Sentencing Commission, Guidelines Manual Supp. to App. C, Amdt. 683 (Nov. 2008) (USSG) (effective Nov. 2006); id., Amdt. 698 (effective Nov. 2007); USSG Supp. to App. C, Amdt. 799 (Nov. 2016); see USSG §1B1.13(b)(5) (Nov. 2025) (providing that “any other circumstance or combination of circumstances that . . . are similar in gravity” to the specifically enumerated ones may constitute “extraordinary and compelling” reasons for a sentence reduction (emphasis added)).

The takeaway from all this is not, as the majority suggests, that compassionate release has always been understood to be limited to certain kinds of factors (i.e., personal circumstances). See ante, at 11–12. Rather, the statutory and regulatory history establishes that §3582(c)(1)(A) provides for flexibility and, with one exception, imposes no substantive categorical restrictions on what may constitute a basis for compassionate release. See §3582(c)(1)(A); see also §994(t) (noting that “[r]ehabilitation” is not on its own a statutorily recognized reason for compassionate release). That view tracks with the language of §3582(c)(1)(A)—which vests district courts with broad discretion to determine whether “extraordinary and compelling reasons” justify a sentence reduction—as well as the unmistakable safety-valve aims of the statute.

2

The majority’s reliance on the habeas-channeling analysis in Preiser and Gonzalez is equally misplaced.

Start with Preiser. In that case, we considered a 42 U. S. C. §1983 claim brought by state prisoners who asserted the unconstitutional deprivation of “good-conduct-time credits” that would have shortened their sentences. 411 U. S., at 476. The prisoners sought “injunctive relief to compel restoration of the credits, which in each case would result in their immediate release from confinement.” Ibid. We held that the state prisoners could not seek such an injunction under §1983 and instead had to proceed under 28 U. S. C. §2254, the habeas statute for state prisoners. Id., at 490, 500. In so holding, we explained that §1983 is a “general” statute that broadly guarantees equitable and legal remedies for the deprivation of any federal right. Id., at 489. Its breadth causes it to bump up against §2254, which is “specific” in that it was “explicitly and historically designed to provide the means for a state prisoner to attack the validity of his confinement” and only after the “exhaustion of adequate state remedies.” Ibid. We feared that privileging the general would cause it to drown the specific. We thus reasoned that Congress’s “specific determination” that “habeas corpus is the appropriate remedy for state prisoners attacking the validity of the fact or length of their confinement” must “override the general terms of §1983.” Id., at 490.

Unlike §1983, §3582(c)(1)(A) is not a “general” statute that protects all rights and provides for all sorts of civil remedies. Instead, it creates a specific mechanism for granting a sentence reduction in the criminal context. Preiser’s reasoning is therefore inapposite, as it would be odd to conclude that Congress intended a statute specifically authorizing release from prison to impliedly limit another statute that authorizes the same. The better inference is that Congress meant for these two statutes to exist side by side and not to the exclusion of one another.[2]

Gonzalez is equally inapposite. There, we held that prisoners cannot use motions for “relief from a final judgment” under Federal Rule of Civil Procedure 60(b) to circumvent the statutory limitations on second-or-successive habeas applications. 545 U. S., at 528. But Rule 60(b), like §1983, is a general mechanism for civil relief. It allows federal courts to “relieve a party . . . from a final judgment” for “any . . . reason that justifies relief.” Fed. Rule Civ. Proc. 60(b)(6). It is a far cry from §3582(c)(1)(A), which, like §2255, specifically authorizes release from prison. Just like Preiser, then, Gonzalez’s interpretation of a general civil provision has nothing to say about §3582(c)(1)(A)’s specific relief scheme and whether Congress meant for it to be cancelled out by an implicit habeas-review bar.

Instead of acknowledging the poor fit of Preiser and Gonzalez, the majority doubles down, insisting that these cases stand for “an anticircumvention principle.” Ante, at 8, n. 3. That might well be a common thread. But the majority does not explain why the anticircumvention principle applies to §3582(c)(1)(A)—a statute that, unlike §1983 or Rule 60(b), specifically authorizes the release of prisoners under given circumstances. For the anticircumvention principle to be apposite, one (specific) statutory directive must be potentially undercut by another (broader) provision. But that does not happen here: Both §2255 and §3582(c)(1)(A) establish specific circumstances under which the release of a prisoner is authorized. So the circumvention concern is not implicated on these facts.

Put differently, the majority assumes that compassionate release operates as an end-run around habeas restrictions. But there is nothing in either statute that corroborates the majority’s intuition that habeas restrictions take primacy over the “extraordinary and compelling reasons” justifying compassionate release.

C

Perhaps even more troubling than the lack of support for today’s holding is the uncertainty this case will create moving forward. That uncertainty results, in part, from the majority’s backwards reasoning: It starts with the premise that Fernandez is functionally attempting to use §3582(c)(1)(A) to attack his conviction collaterally and, from there, concludes that §2255 must prevent this sort of end-run around the strict process for habeas relief. But that reasoning rests on the unfounded assumption that Fernandez is essentially seeking to have his conviction vacated rather than his sentence shortened. Those are two completely different asks. And the majority’s reasoning also merely assumes that they are incompatible.

Worse still, on its way to announcing that “[a] prisoner who collaterally attacks the validity of his conviction must proceed through” habeas only, the majority fails to define with any precision what it means for a prisoner to “collaterally attac[k] the validity of his conviction.” Ante, at 1. Its inability to do so is understandable, since delineating the bounds of such “collateral attacks” is notoriously difficult. Cf. W. Baude, J. Goldsmith, J. Manning, J. Pfander, & A. Tyler, Hart and Wechsler’s The Federal Courts and the Federal System 1687 (8th ed. 2025) (Hart & Wechsler) (discussing the difficulty of determining when, under the Preiser line of cases, a successful §1983 action necessarily implies “the invalidity of the plaintiff ’s conviction”).[3]

This case illustrates why the new line the majority draws is a fuzzy one. The District Court did not question the legality of the jury’s verdict or Fernandez’s conviction. See 2022 WL 17039059, *4 (SDNY, Nov. 17, 2022). To the contrary, it found that despite the lawfulness of Fernandez’s conviction, its lingering “concer[n]” as to his guilt was an “extraordinary and compelling” reason for a sentence reduction. Ibid. A prisoner who proceeds through §2255, however, must show that his sentence was unlawful—specifically, that it was “imposed in violation of the Constitution or laws of the United States.” §2255(a). That is something Fernandez cannot definitively do, given this Court’s expressed doubts about whether “a freestanding claim of actual innocence” suffices to establish the unlawfulness of a sentence. McQuiggin v. Perkins, 569 U.S. 383, 392 (2013).

So, how, exactly, does the District Court’s “extraordinary and compelling” finding qualify as a collateral attack on Fernandez’s conviction for §2255 purposes, as the majority maintains? See ante, at 14. The District Court did not find, for instance, that Fernandez should be released from prison because his sentence or conviction was unconstitutional or otherwise unlawful. Rather, the District Court considered whether Fernandez might actually be innocent—something we have never recognized as a potential ground for habeas relief. See McQuiggen, 569 U. S., at 392. It is hard to see why this consideration actually creates the end-run around habeas the majority fears.[4]

One could easily imagine similar difficulties arising in other cases. Consider, for instance, a federal prisoner who files a §2255 motion arguing that his conviction and sentence are unconstitutional because a chronic mental illness rendered him incompetent to plead guilty or stand trial. In resolving the §2255 motion, the district court might reject as unreliable the evidence supporting the severity of the prisoner’s condition. See, e.g., Cohn v. United States, 2023 WL 4670839, *13 (NDNY, July 10, 2023) (rejecting §2255 motion on these grounds). Under the majority’s rule, would that prisoner be precluded from citing his mental condition in a subsequent §3582(c)(1)(A) motion? Maybe. But that would certainly be an odd result, given the majority’s suggestion that chronic illnesses are “heartland ‘extraordinary and compelling reasons’ that might warrant an early release from prison.” Ante, at 13.

Thus, the line between a proper compassionate-release claim and a proper habeas claim could be difficult to draw moving forward. The majority ignores this problem rather than taking it as a clue that the habeas-based limitation it favors is an ill fit for the distinct compassionate-release evaluation.

III

For the reasons I have given, I think the majority is wrong to hold, as a categorical matter, that conviction-related concerns a prisoner may raise under §2255 can never be grounds for granting a motion under §3582(c)(1)(A). In my view, all facts and circumstances are eligible to be considered as grounds for granting a compassionate-release motion. Other than the rehabilitation restriction, see 28 U. S. C. §994(t), the only statutory limit on the District Court’s discretion is one of degree. So, if the proffered facts and circumstances (of whatever type) provide an “extraordinary and compelling reason” to reduce the prisoner’s sentence (alone or in combination), the District Court may grant compassionate release.

I do not take a position as to whether Fernandez’s case meets that threshold. The District Court thought so—it laid out the evidentiary gaps that caused it to have serious doubts about whether Fernandez was involved in the underlying crime at all. See 2022 WL 17039059, *4.[5] Also notable, the District Court based its §3582(c)(1)(A) ruling on both its “disquiet” about Fernandez’s potential innocence and the significant disparity between Fernandez’s sentence and his codefendants’ sentences. Ibid. But due to the categorical rule the Second Circuit employed, that court did not address whether the District Court erred in concluding that §3582(c)(1)(A)’s “extraordinary and compelling reasons” threshold was satisfied on these facts. See 104 F. 4th 420, 429–433 (2024). I would vacate the judgment below and remand for the Second Circuit to make that assessment in the first instance.[6]

*  *  *

The Court’s holding in this case reflects an unfortunate misunderstanding of the safety-valve mechanism Congress created in 1984 and broadened in 2018. Congress designed compassionate release as a tool for preserving a modicum of mercy in an otherwise harsh sentencing system. And nothing about the text or history of the compassionate-release provision suggests that Congress meant for this discretionary second-look opportunity to be cabined in the way the majority suggests. In other words, Congress gave prisoners an equitable “out” for the extraordinary and compelling circumstance in which they might otherwise be stuck serving an unjust sentence. Ironically, then, the escape hatch the majority decries is actually §3582(c)(1)(A)’s core mission.

Deploying its own concerns about habeas end-runs, the majority misses all this; it now drastically limits the availability of the compassionate-release tool Congress crafted. And it does so without adequately identifying any true conflict with §2255. The majority’s atextual and unsupported limitation on a district court’s sentencing discretion is an unnecessary rewriting of the statute Congress wrote and an unwarranted revision of the compassionate-release scheme Congress intended to establish.