BETTY B. FLETCHER, Circuit Judge:

We review three consolidated appeals that present one overarching issue: Did the district court follow the mandate of Gisbrecht v. Barnhart, 535 U.S. 789, 122 S.Ct. 1817, 152 L.Ed.2d 996 (2002), in determining the amount of attorneys' fees awarded to lawyers who successfully represented Social Security disability insurance ("SSDI") claimants in federal court under contingent-fee contracts? We hold in each case that it did not. We vacate the district courts' orders and grant the attorneys the contingency-based fees they requested.

I.

In each of the three cases presented for review, the Social Security Administration ("SSA") denied a claim for benefits. Each claimant retained an attorney to challenge the administrative action in federal court. Each of the three claimants was represented by a different attorney: Brian C. Shapiro represented Clara Crawford; Young Cho represented Ruby Washington; and Denise Bourgeois Haley represented Daphne M. Trejo. All three attorneys were affiliated with the Lawrence D. Rohlfing ("Rohlfing") law firm, which specializes in Social Security matters. In each case, the claimant signed a written contingent-fee agreement under which the attorney would be paid 25% of any past-due benefits awarded to the claimant. The district court remanded each case back to the SSA, which eventually awarded substantial past-due benefits to each claimant. The attorneys subsequently filed motions in the district court pursuant to 42 U.S.C. § 406(b) requesting fees of less than the 25% of past-due benefits for which their attorney-client fee agreements provided. In each case, the Commissioner of Social Security ("Commissioner") declined to assert a position on the reasonableness of the fees requested by the attorneys. None of the claimants objected to the requested fees. Nevertheless, the district court in each case awarded significantly lower fees than the attorneys sought.

A. Crawford ( Real-party-in-interest Shapiro )

On remand, the SSA awarded Crawford $123,891.20 in past-due benefits. Twenty-five percent of that amount is $30,972.80. Shapiro requested a fee of $21,000 (16.95% of the past-due benefits awarded). To support his motion, Shapiro presented evidence that he spent 19.5 hours of his own time and 4.5 hours of paralegal time on Crawford's case. Because the Rohlfing firm operates only on a contingent-fee basis, Shapiro presented evidence (1) that small firms charged average hourly rates, after accounting for inflation, of $272.72 for partners, $186.54 for associates, and $93.82 for paralegals; (2) that the upperdecile hourly rates charged for each position in small firms were $366.54, $272.72, and $130.91, respectively; and (3) that the average gross income of small plaintiffs' contingency firms was about 30-40% higher than the average gross income of all small law firms combined. He also presented evidence that a lawyer representing SSDI claimants in federal court could anticipate payment in only about 35% of cases because (1) district courts granted benefits in SSDI cases only about 6% of the time, and (2) in the 48% of cases in which the district court remanded an SSDI case to the agency, the SSA awarded benefits about 60% of the time. To aid the district court's evaluation of his request, Shapiro noted that the fee he requested was equivalent to 3.55 times the lodestar calculation and reasonably accounted for the risk he assumed in representing his client on a contingent-fee basis.

The court found that there was no "fraud or overreaching" in the negotiation of the 25% contingent-fee agreement, that the requested fee was within the 25% boundary set by § 406(b), and that Shapiro caused no unnecessary delay resulting in an undue accumulation of back benefits. The court also found that the hours, comparative hourly rates, and inflation rates that Shapiro presented were reasonable and would result in a lodestar fee of $5,907.14. The court concluded, however, that Shapiro had not met his burden of convincing the court that a fee of $21,000, which the court found represented a 256% "enhancement" over the lodestar, was reasonable. The court noted that Shapiro did not present "data regarding his firm's success rate that would enable the Court to assess the risk assumed by his firm in representing social security benefits claimants in the Central District of California" and determined that counsel's expertise had already been taken into account in the comparative hourly rate. The court therefore found that the fee constituted a "windfall" and that a "substantial reduction [was] warranted." The court concluded, without explaining its reasoning, that a 40% percent enhancement over the lodestar fee would be reasonable. The court awarded a fee of $8,270.00, or 6.68% of the past-due benefits.

B. Washington

( Real-party-in-interest Cho )

On remand, the SSA awarded Washington $76,041.00 in past-due benefits. Twenty-five percent of that amount is $19,010.25. Cho sought a fee of $11,500.00 (15.12% of the benefits awarded). Cho presented evidence that he spent 17.45 hours of his time and 4.7 hours of paralegal time on the case. Cho introduced the same comparative hourly rate data that Shapiro presented, as well as evidence that the costs of legal services had risen faster than the cost of living, justifying hourly rates in the ninth decile of $429.95 for partners, $319.90 for associates, and $153.55 for paralegals. Finally, Cho presented the same success-rate data for SSDI cases as did Shapiro. Cho noted that courts determining fees in class-action securities cases had found that multipliers of 3 to 4.5 over the lodestar were reasonable.

The same magistrate judge who decided Crawford also decided Washington. As in Crawford, the court found that there was no "fraud or overreaching" in the making of the contingent-fee agreement, that the fee Cho sought fell within the 25% statutory boundary and was less than the fee Washington agreed to pay, and that there was no "excessive delay" attributable to counsel. The court also found that the economic data Cho presented were reasonable. Nevertheless, the court found that the requested fee, which it found represented an 82% "enhancement" over the $6,303.95 lodestar fee, was unreasonable. As in Crawford, the court noted that Cho had provided no data regarding his law firm's success rate and that Cho's skills were already accounted for in determining the reasonable hourly rate. The court also found that: (1) "counsel did not have to do much work to persuade the Commissioner to stipulate to a remand," and (2) a fee representing a 3 to 4.5 multiplier over the lodestar would be unreasonable because Social Security cases are not as complex or expensive as class-action securities litigation. Concluding that the requested fee would represent a "windfall," the court applied a 40% enhancement to the lodestar fee, without explaining why that percentage increase would result in a reasonable fee. The court awarded a fee of $8,825.53, or 11.61% of the past-due benefits.

C. Trejo ( Real-party-in-interest Haley )

On remand, the SSA awarded Trejo $172,223.00 in past-due benefits. Twenty-five percent of that amount is $43,055.75. Haley sought a fee of $24,000 (13.94% of the past-due benefits awarded.) She presented evidence that she spent 26.9 hours of her time and 2.6 hours of paralegal time on the case. She also presented the same comparative hourly rate data and success rate data as did Shapiro and Cho.

The court acknowledged the contingent-fee agreement providing for 25% of the past-due benefits awarded and found that there was no evidence of fraud or overreaching by counsel in making the agreement. The court noted the "high quality of the representation provided . . . which ultimately resulted in a fully favorable decision awarding over twelve years of back benefits" and found that there was no excessive delay attributable to Haley. The court also found that Haley's economic data were reasonable. Nevertheless, the court found that the fee Haley sought was unreasonable. First, finding that 1.5 hours of claimed paralegal time and 1.4 hours of claimed attorney time were improperly attributed to the federal court action, the court reduced the hours worked to 25.5 hours of attorney time and 1.1 hours of paralegal time. Second, as in Crawford and Washington, the court noted that Haley had provided no data regarding her law firm's success rate and that Haley's skills were already accounted for in determining the reasonable hourly rate. Third, the court found that there was no evidence that the firm was precluded from other employment by accepting Trejo's case and that the case did not involve any unduly short time limits. The court concluded that "plaintiff's counsel has done a wholly inadequate job convincing the Court that the 279% enhancement sought . . . is reasonable under the circumstances presented," and that a reasonable enhancement over the $6,325.20 lodestar would be 100%. The court did not, however, explain why a 100% enhancement would be reasonable. The court determined that a reasonable fee would be $12,650.40, or about 7.35% of the past-due benefits.

II.

The three attorneys timely appealed the district courts' fee orders. We consolidated the cases for argument. We review a district court's award of attorneys' fees pursuant to 42 U.S.C. § 406(b) for abuse of discretion. Clark v. Astrue, 529 F.3d 1211, 1213 (9th Cir. 2008) (citing Allen v. Shalala, 48 F.3d 456, 457 (9th Cir. 1995), abrogated on other grounds by Gisbrecht, 535 U.S. at 799, 122 S.Ct. 1817). "The district court abuses its discretion if it does not apply the correct legal standard or rests its decision on a clearly erroneous finding of fact." Id. at 1214. We review de novo the district court's interpretation of a statute. Id.; see also Mudd v. Barnhart, 418 F.3d 424, 427 (4th Cir. 2005) (reviewing de novo the district court's interpretation of 42 U.S.C. § 406(b)).

A.

Under 42 U.S.C. § 406(b), a court entering judgment in favor of an SSDI claimant who was represented by an attorney "may determine and allow as part of its judgment a reasonable fee for such representation, not in excess of 25 percent of the total of the past-due benefits to which the claimant is entitled by reason of such judgment." In contrast to fees awarded under fee-shifting provisions such as 42 U.S.C. § 1988, the fee is paid by the claimant out of the past-due benefits awarded; the losing party is not responsible for payment. Gisbrecht, 535 U.S. at 802, 122 S.Ct. 1817. Also in contrast to fees awarded under fee-shifting statutes, under which "nothing prevents the attorney for the prevailing party from gaining additional fees, pursuant to contract, from his own client," id. at 806, 122 S.Ct. 1817, the court-awarded fee is the only way a successful SSDI attorney may recover fees for work performed before the district court. In fact, it is a criminal offense for an attorney to collect fees in excess of those allowed by the court. 42 U.S.C. § 406(b)(2); see also Gisbrecht, 535 U.S. at 806-07, 122 S.Ct. 1817.

SSDI attorneys routinely enter into contingent-fee agreements specifying that the fee will be 25% of any past-due benefits recovered, thus providing the attorney the statutory maximum of fees if the representation is successful. See Gisbrecht, 535 U.S. at 803, 122 S.Ct. 1817. Contingent-fee agreements were already prevalent at the time § 406(b) was enacted, see id. at 805, 122 S.Ct. 1817, and agreements providing for fees of 25% of past-due benefits have since become the "most common fee arrangement between attorneys and Social Security claimants." Id. at 800, 122 S.Ct. 1817. The statute does not specify how courts should determine whether a requested fee is reasonable. See id. (noting "the statute's inconclusive text"). Rather, the statute provides only that the fee must not exceed 25% of the past-due benefits awarded.

Before Gisbrecht, the circuits were split regarding how to address this ambiguity in the statute. We belonged to the majority of circuits which used the lodestar method to determine a reasonable fee under § 406(b). See Gisbrecht v. Apfel, 238 F.3d 1196, 1197-98 (9th Cir. 2000); see also Allen, 48 F.3d at 458. Under the lodestar method, the district court determines a reasonable fee by multiplying the reasonable hourly rate by the number of hours reasonably expended on the case. Gisbrecht, 238 F.3d at 1197-98. Although the district court could consider the contingent nature of the fee agreement in determining a reasonable fee, we held that failure to do so was not an abuse of discretion. See id. at 1199. Three other circuits rejected the lodestar method, instead giving effect to the attorney-client contingent-fee agreement unless the resulting fee was unreasonable. See Wells v. Sullivan, 907 F.2d 367, 370 (2d Cir. 1990); Rodriquez v. Bowen, 865 F.2d 739, 746 (6th Cir. 1989) (en banc); McGuire v. Sullivan, 873 F.2d 974, 980-81 (7th Cir. 1989); see also Gisbrecht, 535 U.S. at 799, 122 S.Ct. 1817 (listing cases).

In Gisbrecht, the Supreme Court flatly rejected our lodestar approach. The Court explained that we had "erroneously read § 406(b) to override customary attorney-client contingent-fee agreements" when we approved the use of the lodestar to determine a reasonable fee, Gisbrecht, 535 U.S. at 808-09, 122 S.Ct. 1817. The Court held that a district court charged with determining a reasonable fee award under § 406(b)(1)(A) must respect "the primacy of lawful attorney-client fee agreements," id. at 793, 122 S.Ct. 1817, "looking first to the contingent-fee agreement, then testing it for reasonableness," id. at 808, 122 S.Ct. 1817. The Court noted that courts that had followed this model had "appropriately reduced the attorney's recovery based on the character of the representation and the results the representative achieved." Id. A fee resulting from a contingent-fee agreement is unreasonable, and thus subject to reduction by the court, if the attorney provided substandard representation or engaged in dilatory conduct in order to increase the accrued amount of past-due benefits, or if the "benefits are large in comparison to the amount of time counsel spent on the case." Id. "[A]s an aid to the court's assessment of the reasonableness of the fee yielded by the fee agreement," but "not as a basis for satellite litigation," the court may require counsel to provide a record of the hours worked and counsel's regular hourly billing charge for noncontingent cases. Id. The attorney bears the burden of establishing that the fee sought is reasonable. Id. at 807, 122 S.Ct. 1817.

The Supreme Court made clear in Gisbrecht why courts must start with the contingent-fee agreement in SSDI cases. The Court explained that the lodestar method was developed to implement fee-shifting statutes, which assess fees against the losing party and which do not prevent the attorney from seeking additional fees from the client. See Gisbrecht, 535 U.S. at 802, 806, 122 S.Ct. 1817. SSDI attorneys' fees, in contrast, are not shifted. They are paid from the award of past-due benefits and the amount of the fee, up to 25% of past-due benefits, is based on the agreement between the attorney and the client. See id. at 803-04, 122 S.Ct. 1817. Congress chose to embrace the use of contingent fees to compensate lawyers who represent SSDI claimants. When Congress added § 406(b) to the Social Security Act in 1965, contingent fees paid from past-due benefits were already the norm. See id. at 804, 122 S.Ct. 1817 (discussing the 1965 amendments' legislative history). Rather than displace contingent-fee agreements, the amendment was meant to address a particular problem in their implementation. Attorneys were entering into agreements reserving what Congress regarded as "exorbitant fees" of one-third to one-half of the past-due benefits, often for easy cases in which the attorney would purposefully cause delay in order to increase the accrued benefits and the resulting fees. See id. At the same time, Congress was mindful that fee awards should be sufficient to encourage adequate representation of claimants. See id. at 806, 122 S.Ct. 1817; see also Wells, 907 F.2d at 370 (contingent-fee agreements "effectuate [C]ongress's objective of securing adequate representation for social security claimants"). Against this backdrop, Congress enacted § 406(b), which says nothing to discourage contingent-fee agreements, instead opting for a 25% cap on fees. See Gisbrecht, 535 U.S. at 805, 122 S.Ct. 1817.

As this history demonstrates, Congress and the Supreme Court have considered the negative policy implications of allowing the lodestar methodology to drive SSDI fee awards. Given the prevalence of lodestar calculations in the fee-shifting context, district courts are familiar with the normal lodestar rules. These include the "strong presumption" that the lodestar is the reasonable fee, see City of Burlington v. Dague, 505 U.S. 557, 562, 112 S.Ct. 2638, 120 L.Ed.2d 449 (1992), and the rule forbidding contingency enhancements to compensate for the risk of non-payment in fee shifting cases. See id. at 566-67, 112 S.Ct. 2638. These rules should not apply where the fee is paid by the client under the agreement negotiated between the parties. The lodestar method under-compensates attorneys for the risk they assume in representing SSDI claimants and ordinarily produces remarkably smaller fees than would be produced by starting with the contingent-fee agreement. A district court's use of the lodestar to determine a reasonable fee thus ultimately works to the disadvantage of SSDI claimants who need counsel to recover any past-due benefits at all.

B.

The Supreme Court in Gisbrecht has given the district courts direction in how to evaluate a request for a contingent fee under § 406(b). Courts must "approach[§ 406(b)] fee determinations by looking first to the contingent-fee agreement, then testing it for reasonableness." Gisbrecht, 535 U.S. at 808, 122 S.Ct. 1817; see also Mudd, 418 F.3d at 428 (recognizing this same methodology). Because the SSA has no direct interest in how much of the award goes to counsel and how much to the disabled person, the district court has an affirmative duty to assure that the reasonableness of the fee is established. Performance of that duty must begin, under Gisbrecht, with the fee agreement, and the question is whether the amount need be reduced, not whether the loadstar amount should be enhanced. An examination of the fee awards in these cases makes it starkly evident that the district courts did not follow this path. Instead, in direct contrast to Gisbrecht's mandate, the district courts' decisions "rest[ed] on lodestar calculations and reject[ed] the primacy of lawful attorney-client fee agreements." Gisbrecht, 535 U.S. at 793, 122 S.Ct. 1817.

Here, in each case, the client signed a contract providing that the attorney would receive 25% of any past-due benefits awarded following appeal to the district court. In determining a reasonable fee, the district courts proceeded contrary to the requirements of Gisbrecht by beginning with a lodestar calculation. Each court compared the lodestar fee to the requested fee award, and referred to the difference between the two as "an enhancement." In each order, the district court expressed the "enhancement" as a percentage of the lodestar for example, the Crawford court noted that the requested fee would have represented a 256% enhancement over the lodestar — and found that each enhancement would be a "windfall" to the lawyers. Each court then added a percentage of the lodestar to enhance the lodestar fees, choosing 40% for the Crawford and Washington cases and 100% for the Trejo case. Had the district court awarded the full fees contracted by the parties, the attorneys in these cases would have received fees ranging from $19,010.25 to $43,055.75. Instead, they received amounts ranging from $8,270.00 to $12,650.40. The fees awarded ranged from 6.68% to 11.61% of the past-due benefits awarded. Put another way, the district courts reduced the contractual fees by between 53.57% and 73.30%.

By beginning with the lodestar calculation, the district courts plainly failed to respect the "primacy of lawful attorney-client fee agreements." Gisbrecht, 535 U.S. at 793, 122 S.Ct. 1817. Lawful attorney-client contingent fee agreements do not result in "enhancements" that modify a lodestar fee that might otherwise be too low. Rather, they are the "primary means" by which fees are determined. Id. at 807, 122 S.Ct. 1817. The district courts' methodology in these cases under-emphasizes the contingent-fee agreements, contravening the Supreme Court's instruction that the agreements be the primary means for determining the fee.

These cases vividly demonstrate the deleterious effect of a district court's failure to recognize the distinction between fee-shifting cases and cases involving payment by the claimant from his benefit award. All of the normal lodestar language and methodology are accounted for in the district court's orders: reasonable fees, reasonable hours, and an "enhancement." Lodestar fees will generally be much less than contingent fees because the lodestar method tends to under-compensate attorneys for the risk they undertook in representing their clients and does not account for the fact that the statute limits attorneys' fees to a percentage of past-due benefits and allows no recovery from future benefits, which may far exceed the past-due benefits awarded. In Crawford, for example, the district court awarded 6.68% of the past-due benefits. From the lodestar point of view, this was a premium of 40% over the lodestar. It seems reasonable. But from the contingent-fee point of view, 6.68% of past-due benefits was over 73% less than the contracted fee and over 60% less than the discounted fee the attorney requested. Had the district court started with the contingent-fee agreement, ending with a 6.68% fee would be a striking reduction from the parties' fee agreement. This difference underscores the practical importance of starting with the contingent-fee agreement and not just viewing it as an enhancement.

The district court orders quote extensively from Gisbrecht. They even cursorily discuss the character of the representation — noting that it was skillful and not dilatory — before concluding that the requested fee would represent a windfall to the attorneys. But this parroting of language from Gisbrecht does not mean that the district courts actually applied its teachings. As the orders make clear, the district courts in these cases started with the lodestar calculation and then adjusted upward to account for the contingent nature of the representation. This is contrary to the Supreme Court's clear directive that the district court must first look to the fee agreement and then adjust downward if the attorney provided substandard representation or delayed the case, or if the requested fee would result in a windfall. See id. at 808, 122 S.Ct. 1817; see also Rodriquez, 865 F.2d at 746 ("In the event the court chooses not to give effect to the terms of the agreement, it should state for the record the deductions being made and the reasons therefore.").

C.

The fees requested here are reasonable under the test mandated by Gisbrecht. Although Gisbrecht did not provide a definitive list of factors that should be considered in determining whether a fee is reasonable or how those factors should be weighed, the Court directed the lower courts to consider "the character of the representation and the results the representative achieved." Gisbrecht, 535 U.S. at 808, 122 S.Ct. 1817; see also Mudd, 418 F.3d at 428 ("The Court did not provide a definitive list of factors to be considered because it recognized that the judges of our district courts are accustomed to making reasonableness determinations in a wide variety of contexts." (quotation and alterations omitted)). The court may properly reduce the fee for substandard performance, delay, or benefits that are not in proportion to the time spent on the case. Gisbrecht, 535 U.S. at 808, 122 S.Ct. 1817. As evidence of the reasonableness of the resulting fee, the court may require counsel to submit a record of hours spent and a statement of normal hourly billing charges. Id. The Supreme Court did acknowledge that the district court could consider the lodestar calculation, but only as an aid in assessing the reasonableness of the fee. See id.

Applying Gisbrecht's reasonableness test, we hold that the attorneys in these cases met their burden to demonstrate that the fees they requested were reasonable. First, no reduction in fees due to substandard performance was warranted. Nothing in the district courts' opinions suggests that counsels' performance was anything other than excellent. In each case, the attorneys demonstrated (and the district courts found) that there was no evidence of fraud or overreaching in the making of the 25% contingent-fee agreements and that they provided high-quality representation which resulted in their clients receiving substantial past-due benefits. Second, no reduction in fees for dilatory conduct was warranted, as the attorneys in these cases caused no excessive delay which resulted in an undue accumulation of past-due benefits. Finally, the requested fees, which were significantly lower than the fees bargained for in the contingent-fee agreements, were not excessively large in relation to the benefits achieved. In each case, counsel voluntarily evaluated the fees in comparison to the amount of time spent on the case. In each case, counsel voluntarily reduced those fees substantially from the allowable 25%. The attorneys will receive no percentage of the substantial future benefits paid to the claimants following their successful representation. The attorneys assumed significant risk in accepting these cases, including the risk that no benefits would be awarded or that there would be a long court or administrative delay in resolving the cases.

Counsel have waited a long, long time for payment, and have borne the costs of this appeal out of the fees to which they are entitled. To force petitioners' counsel to multiply the diminution of those fees by going through three separate and unnecessary district court proceedings would be completely unjustified. The question before the district courts was whether the fees sought by petitioners are reasonable. See Gisbrecht, 535 U.S. at 809, 122 S.Ct. 1817 (holding that 42 U.S.C. § 406(b) "instructs courts to review for reasonableness fees yielded by [attorney-client contingent-fee] agreements."); see also Wells, 907 F.2d at 371 ("[B]ecause a successful social security claimant evaluates and pays his own attorney, a court's primary focus should be on the reasonableness of the contingency agreement in the context of a particular case."). Because we have held that the fees are reasonable, nothing remains for the district courts to do in these cases, except to award those fees. The Supreme Court in Gisbrecht held that "satellite litigation" over attorneys' fees should not be encouraged. See Gisbrecht, 535 U.S. at 808, 122 S.Ct. 1817. Accordingly, we hold that petitioners are entitled to the fees which they requested and remand with instructions to award the requested fees.

D.

A separate and adequate ground for vacating the orders in these cases is the district courts' failure to explain why the percentages by which they enhanced their lodestar calculations produced a reasonable fee in each case. Although the district court has discretion to determine a reasonable fee, it must provide "a concise but clear explanation of its reasons for the fee award." Hensley v. Eckerhart, 461 U.S. 424, 437, 103 S.Ct. 1933, 76 L.Ed.2d 40 (1983). Here, although each district court explained on a very general level that the requested fee would result in a windfall to the attorney, each court failed to relate its "enhancement" of the lodestar to the circumstances of the individual case. See, e.g., Moreno v. City of Sacramento, 534 F.3d 1106, 1112 (9th Cir. 2008) (noting that where the district court awards a substantially reduced fee, it must "articulate[] its reasoning with more specificity"); Gates v. Deukmejian, 987 F.2d 1392, 1400 (9th Cir. 1992) (as amended) ("the use of percentages" does not "discharge[] the district court from its responsibility to set forth a `concise but clear' explanation of its reasons for choosing a given percentage reduction").

In addition, the district courts' orders in Trejo and Crawford misconstrue the nature of the risk assessment by focusing on the firm's overall success rate instead of the specific facts that make a given case more or less risky for the firm. For example, in Crawford, the district court faulted the firm for failing to "provide[] any data regarding [the] firm's success rate that would enable the Court to assess the risk assumed by [the] firm in representing social security benefits claimants in the Central District of California." This misstates the attorney's burden, which is to show that the fee is reasonable based on the facts of the particular case. A district court cannot reduce the amount of a fee simply because a firm is generally successful. Rather, the district court should look at the complexity and risk involved in the specific case at issue to determine how much risk the firm assumed in taking the case.

III.

Because the district courts inverted the reasonableness analysis prescribed by Gisbrecht, and because the attorneys proved the reasonableness of their fees and none of the grounds in the district court orders establish unreasonableness, we direct the district court to grant the attorneys their requested fees. We VACATE the district courts' orders and REMAND with instructions to order the SSA to pay the attorneys the fees they requested, less fees already paid under the EAJA, and to release the balance of the withheld past-due benefits to the claimants.

VACATED and REMANDED WITH INSTRUCTIONS.