Why this result mattered
Functional tie
The principal comparison placed Qwen 3.8 and Claude under concealed A/B labels.
Both answers direct StreamLedger to seek prompt, targeted production and limited deferral while preparing and timely filing a present-record opposition, rejecting a broad six-month stay and preserving narrower subclass or issue-class arguments. Their differences in emphasis, scheduling, and internal preparation do not demonstrate a different filing outcome or concrete consequence to StreamLedger.
Fast audit
Open the underlying record
These links open all four answers, the principal blind judgment, original prompt, and sealed key.
Prompt and sealed key
The key is an audit reference. It was not shown to the contestant systems and did not determine the comparative verdict.
Read the original matter prompt
Question 57: The Subscription Fee Class Before the Files Arrive
Your role
You represent StreamLedger, an online bookkeeping company, in opposing a proposed consumer class concerning a $7.95 priority-processing fee. Plaintiffs moved for class certification before producing the named plaintiffs' account files. StreamLedger must decide what to file now and whether to seek a discovery order before briefing the merits.
Governing law
Assume the following class-action rules apply:
A class may be certified only if the proposed representatives have standing and satisfy numerosity, commonality, typicality, and adequacy.
For a damages class, common questions must predominate over questions affecting only individual members, and a class action must be superior to feasible alternatives.
A uniform material misrepresentation may support common proof. Materially different disclosures, enrollment paths, oral communications, or assent records may require individualized proof or separately defined subclasses.
The consumer statute requires that each claimant was exposed to a deceptive representation or omission, that the deception caused the challenged payment, and that the claimant suffered an unreimbursed economic loss. Reliance may be inferred only from a materially uniform representation.
A contract claim depends on the terms presented and assented to by the customer. A customer cannot recover a fee that the customer knowingly and separately selected under an enforceable disclosed term.
A damages model must match the certified theory. Full refunds, chargebacks, and employer reimbursements must be deducted from any out-of-pocket model.
A named plaintiff's standing and typicality must be established from evidence, not allegations alone. Material uncertainty about the representative's enrollment path, assent, refund status, or personal loss may defeat or postpone certification.
The court may defer certification briefing and compel targeted discovery when the movant controls specific, previously requested information essential to standing, typicality, or predominance. The requesting party must identify the documents and explain their decisional significance; broad delay for unfocused discovery is not permitted.
A defendant opposing certification must connect variations to elements or remedies. Merely asserting that every customer is different does not defeat predominance.
The court may certify a narrower issue class or subclasses when genuinely common issues can be separated without making the proceeding unmanageable.
Facts
Between January 2023 and December 2025, StreamLedger charged 62,400 customers a $7.95 priority-processing fee. Its records identify four enrollment paths:
- 14,800 desktop customers saw a separate unchecked box stating, “Add priority verification for $7.95; ordinary verification remains available at no charge.”
- 21,300 mobile customers saw a preselected box stating, “Priority processing, $7.95,” with a link explaining the free ordinary option.
- 17,100 legacy customers enrolled through an email link whose landing page disclosed the fee only in the linked service terms.
- 9,200 customers enrolled through customer-service calls using six different scripts. StreamLedger retained only a sample of recordings from each script period.
Backend records show that priority processing reduced average verification time by 18 hours on weekdays but produced no average improvement on weekends. The complaint alleges that the fee was uniformly hidden and provided no benefit.
StreamLedger issued 5,400 full fee refunds and processed 1,180 chargebacks. Its business-account field identifies 7,900 payments made on employer cards, but that field does not establish whether the customer later reimbursed the employer or personally bore the expense.
The proposed class is every person charged the fee. Plaintiffs assert consumer-statute and contract claims. Their certification motion relies on one marketing page and the common backend queue.
The three proposed representatives are:
- Milo Chen, who alleges he enrolled on mobile but testified that he does not remember whether the fee box was already selected or whether he opened the linked explanation.
- Tara Vale, who alleges a call-center representative said the fee was mandatory. She does not remember the date or representative.
- Jae Morgan, who alleges no refund was received. StreamLedger's ledger contains an ambiguous “CB-R” code that can mean either a completed chargeback or a rejected request.
Four months ago StreamLedger requested each named plaintiff's original confirmation emails, screenshots, bank or card statements, refund or chargeback records, employer-reimbursement records, and communications concerning enrollment. Plaintiffs agreed to produce the files by June 1. They produced none of them and filed for certification on June 3. Their opposition brief is due in 21 days.
Plaintiffs argue that the missing documents are merits details and that StreamLedger's own database should be enough. StreamLedger wants either immediate denial, a broad six-month discovery stay, or full merits briefing now.
Assignment
Analyze:
- Standing, typicality, and adequacy of each proposed representative on the present record.
- Which liability questions are common and which require proof tied to enrollment path.
- Predominance for the consumer-statute and contract claims.
- The effect of the four disclosure paths, six call scripts, weekday and weekend performance, and differing assent records.
- Refunds, chargebacks, employer payments, and the required damages model.
- Whether an issue class or subclasses could cure some variations.
- The exact named-plaintiff materials that matter and why.
- Whether StreamLedger should move to defer and compel targeted production, oppose certification now, request broad discovery, or combine those approaches.
- The concrete filing sequence for the next 21 days without overstating what the missing files will prove.
Read the sealed key
Q57 key (Subscription Fee Class)
- StreamLedger should move immediately to defer the certification response and compel the specifically promised named-plaintiff files, while preserving a short merits opposition based on the existing record. The agreed production is two days overdue and directly bears on standing and typicality. Missed: the court decides representation on allegations when the decisive personal records remain in plaintiffs' control.
- The targeted materials are confirmation emails, screenshots, bank or card statements, refund and chargeback records, employer-reimbursement records, and enrollment communications. They test Milo's disclosure path and assent, Tara's alleged oral representation, and Jae's personal loss and ambiguous chargeback. Missed: a broad delay request fails because it does not connect discovery to elements.
- Common issues exist, including the backend queue's operation and whether particular standardized screens were misleading. They do not establish uniform exposure or causation across four enrollment paths, six scripts, and weekday-versus-weekend performance. Missed: one common system fact is treated as predominant liability proof.
- Predominance is weak for one undivided consumer class. The statute requires exposure, causation, and unreimbursed loss; only a materially uniform representation permits inferred reliance. Desktop opt-in, mobile preselection, linked legacy terms, and variable calls require path-specific analysis. Missed: customers who saw materially different propositions are adjudicated as one transaction.
- The contract claim also turns on the term presented and assented to. The disclosed desktop group may have knowingly purchased a separate service, while legacy and call-center customers present different formation records. Missed: contract liability is assumed from a common charge code rather than common terms.
- Any damages model must remove 5,400 refunds, 1,180 completed chargebacks once identified, and employer-funded amounts that did not become personal loss. The “CB-R” code and employer-card field require record resolution, not assumptions in either direction. Missed: recovery includes people who lost nothing or excludes people who did.
- Narrow subclasses by enrollment path and perhaps weekday or weekend performance may be possible after discovery; an issue class on backend operation may also be considered, but neither cures named-plaintiff standing or path representation. Missed: subclassing is offered as a label without a representative and matched theory.
- Sequence: file a focused motion to defer and compel now, attach the four-month-old requests and June 1 agreement, propose expedited production and limited depositions, and submit a protective opposition identifying present predominance defects without claiming the missing files will necessarily defeat certification. Missed: StreamLedger either waives its developed opposition or seeks an unjustified six-month pause.
Direct judge report
The report below is preserved from the Qwen 3.8 versus Claude verdict record.
Read the preserved judge report
Basis. Both answers direct StreamLedger to seek prompt, targeted production and limited deferral while preparing and timely filing a present-record opposition, rejecting a broad six-month stay and preserving narrower subclass or issue-class arguments. Their differences in emphasis, scheduling, and internal preparation do not demonstrate a different filing outcome or concrete consequence to StreamLedger.
[B] rejected: “terms versions across 36 months” — record lacks evidence of changing terms versions.
[B] rejected: StreamLedger “has already run” a claims-administered refund program — record establishes refunds, not such a program.
[A] rejected: mobile logs may show whether Chen opened the linked explanation — record lacks evidence that link-access logs exist.
[A] rejected: plaintiffs should produce an explanation of StreamLedger’s “CB-R” code — internally inconsistent with the code being in StreamLedger’s ledger and with the proposed self-production of corresponding records.
Failed candidates. Denial-first versus deferral-first framing — different conduct failed because both answers recommend filing the targeted motion and preserving a timely opposition seeking present-record denial.
Failed candidates. B’s additional internal evidence-building steps — concrete consequence failed because the record does not show that A’s reader would omit ordinary preparation or lose any argument.
Failed candidates. Ten-to-fourteen-day production schedule versus a 45–60-day deferral — concrete consequence failed because both preserve the 21-day deadline and no deadline loss or strategic harm is established.
Failed candidates. Different issue-class formulations — likelihood failed because both treat subclasses or issue certification only as a narrow fallback after representative and evidentiary defects are addressed.
Failed candidates. Unsupported terms-version, refund-program, and link-log assertions — no safe neutralization failed because ordinary record verification rejects them without changing the recommended course.
Hard stop. The sentence cannot be completed with a concrete, record-supported consequence.
All six pairings
Every pair of the four model answers was judged independently under the same evaluation instrument.
Open the six head-to-head results
| Pairing | Result | Why it mattered | Record |
|---|---|---|---|
| Qwen 3.8 27B vs Claude Opus 5 | Functional tie | Both answers direct StreamLedger to seek prompt, targeted production and limited deferral while preparing and timely filing a present-record opposition, rejecting a broad six-month stay and preserving narrower subclass or issue-class arguments. | Judge record |
| Qwen 3.8 27B vs DeepSeek V4-Flash-0731 | Functional tie | Both answers direct StreamLedger to seek limited deferral and targeted named-plaintiff production while preserving a timely present-record opposition, and both reject a broad six-month stay. | Judge record |
| Qwen 3.8 27B vs Qwen 3.5 27B | Functional tie | Both answers place StreamLedger on the same material path: promptly seek targeted named-plaintiff production, reject a broad six-month stay, preserve a timely present-record certification opposition, and request supplemental consideration after production. | Judge record |
| DeepSeek V4-Flash-0731 vs Claude Opus 5 | Functional tie | Both answers put StreamLedger on the same material path: promptly seek targeted production and deferral, file a timely alternative opposition challenging the representatives, pathwide predominance, and damages model, and reject a broad six-month stay. | Judge record |
| DeepSeek V4-Flash-0731 vs Qwen 3.5 27B | Functional tie | Both answers direct StreamLedger to seek targeted production of the named plaintiffs’ files, oppose certification on the present record, reject a broad stay, and preserve path-specific predominance, assent, and damages objections. | Judge record |
| Qwen 3.5 27B vs Claude Opus 5 | Claude win | Answer A directs StreamLedger to use the remaining 21 days to create and file an authenticated evidentiary record supporting its certification opposition even if deferral is denied. | Judge record |
Rejected-point diagnostic
Judge-attributed rejected points for this matter across all six pairings:
| Qwen 3.8 27B | 2 |
|---|---|
| DeepSeek V4-Flash-0731 | 2 |
| Claude Opus 5 | 7 |
| Qwen 3.5 27B | 10 |
This is a review-burden diagnostic, not a universal hallucination measure.