Price comparison only·Not medical advice·Using this site means you accept our Terms & Privacy·Always confirm with your pharmacist·Price comparison only·Not medical advice·Using this site means you accept our Terms & Privacy·Always confirm with your pharmacist·
Commissioners Court backup for the award of RFP 1311-003-CW, pharmacy benefit management services, to Envision Pharmaceutical Services
Pharmacy benefit manager: Envision Pharmaceutical Services, LLC. Term October 1, 2014 to September 30, 2017. Item 21 on the June 10, 2014 Commissioners Court agenda. Three year term effective October 1, 2014. The backup includes AWP discount tables, dispensing fees, rebate and guarantee language. No current Travis County PBM contract is posted.
No page of the posted agenda backup is redacted. This is the Commissioners Court award packet, not the executed agreement; the contract itself is not posted.
The rubric
Clause by clause
Each item quotes the contract and cites the page it came from. Items whose pages are redacted are marked unclear and left out of the score.
Fiduciary duty and loyalty
Weight 7 of 100, model section Section 2.4
Not addressed
All five models reached this verdict. The agenda request on pages 1 to 3, the pharmacy memorandum on pages 8 to 11 and the consultant notes on pages 16 to 18 describe a transparent business model and a fee only compensation promise, but no passage makes Envision owe a duty of loyalty or care to the county or its members, and there is no disgorgement remedy. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
The PBM owes duties of loyalty and care solely to the plan and its members, must disclose all direct and indirect compensation and every economic conflict, and must disgorge anything retained in breach without the plan proving damages.
Without a loyalty duty the PBM is free to put its own affiliates and revenue ahead of the plan when it exercises discretion over formularies, networks and pricing.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Full disclosure of compensation and conflicts
Weight 6 of 100, model section Sections 2.4, 2.6
Weaker than the model
Three of five models read this as partial; Gemini 3.1 Pro and Grok 4.6 called it missing because a single fee promise is not a disclosure covenant. The consultant records on page 17 that Envision will put in writing that it accepts only its per employee per month fee and no other money tied to the county's drug spend, and page 11 prints that fee as $3.50. That is a real statement about compensation, but there is no schedule of direct and indirect compensation, no affiliate or consultant payment disclosure and no officer certification on any cycle. Page 10 also warns that additional fees for services are in the contract that is not attached, so the packet addresses the subject without meeting the model.
Envision will put in writing that they will only accept their per employee/per month fee and not accept any other monies associated with the prescription drug spend of Travis County.
Page 17Model contract says
All PBM and related-entity compensation, affiliate ownership touching plan claims, and payments to consultants or brokers are disclosed, and a named officer certifies quarterly and annually that everything has been disclosed and remitted.
Hidden revenue streams are the main way plans overpay. Officer certification puts a name on the line.
Model votes: Kimi K3 partial, Grok 4.6 missing, GPT 5.6 Sol partial, Claude Opus 5 partial, Gemini 3.1 Pro missing
Ownership and affiliate disclosure exhibit
Weight 3 of 100, model section Proposed Exhibit A-5
Not addressed
All five models reached this verdict. No page contains a signed list of owners, parents, subsidiaries, group purchasing organizations, rebate aggregators or owned pharmacies, and nothing requires Envision to refresh such a list. The only corporate relationships named are the Costco partnership for mail and specialty on page 10 and the note on pages 4 and 8 that the outgoing vendor OptumRx is a United Healthcare subsidiary. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
A signed exhibit lists every owner, parent, subsidiary, affiliate, group purchasing organization, rebate aggregator, mail order pharmacy and specialty pharmacy tied to the PBM, and the PBM must refresh it within 30 days of any change in ownership or control.
The model contract has no ownership exhibit, and the Nautilus Health Institute review of version 3.5 scored conflict of interest 70 of 100 for that gap. A plan cannot police affiliate pricing, steering or GPO pass through if it does not know which companies are affiliates.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Definitions that close loopholes
Weight 5 of 100, model section Section 1
Not addressed
All five models reached this verdict. Rebate, pharmacy discounts, mail order, specialty and administrative fee are all used in the pricing summary on pages 10 and 11 and in the narrative on pages 2 and 17, but none is defined anywhere. Without a definitions article there is no way to tell whether the 100 percent pass through promise reaches administrative fees, data fees or other manufacturer payments. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
Rebate, manufacturer revenue, affiliate, specialty, generic and pharmacy are defined by economic function, not by label, so value cannot be relabeled (for example as fees) to escape pass-through or guarantees.
Most PBM revenue leakage happens inside definitions. A narrow definition of rebate lets fees, grants and administrative payments stay with the PBM.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
100% pass-through of manufacturer revenue
Weight 11 of 100, model section Section 4
Weaker than the model
Four of five models read this as partial; Kimi K3 called it good because the page 11 table adds rebate minimums of $17.35 per Brand Rx at retail and $66.60 per Brand Rx at mail to the 100 percent pass through label. The county's memorandum on page 10 says Envision has agreed in contracts to pass through 100 percent of all monies received or earned through county utilization, and page 17 adds that rebates arrive at the point of sale rather than after a year end reconciliation. The promise is broad on its face, but the packet gives no definition of the money covered, no payment schedule, no closed list of excluded claims and no audit right, and the operative contract is not in the file. A summary of a proposal term cannot be read as the enforceable clause the model contract published by Mark Cuban and collaborators requires, so the majority verdict stands.
agreed in contracts to pass through 100% of all monies received or earned through the utilization of Travis County and its health plan members.
Page 10Model contract says
One hundred percent of manufacturer revenue of every kind is paid to the plan, quarterly within 30 days of quarter end, with only a closed list of excludable claims.
Rebates and other manufacturer payments are the largest dollar item in most PBM contracts. Anything less than 100% of everything is money left with the PBM.
Model votes: Kimi K3 good, Grok 4.6 partial, GPT 5.6 Sol partial, Claude Opus 5 partial, Gemini 3.1 Pro partial
Affiliate pharmacy pricing at acquisition cost
Weight 5 of 100, model section Section 3.2
Not addressed
Four of five models read this as missing; GPT 5.6 Sol called it partial because the no spread sentence on page 10 reaches mail and specialty. Page 10 describes a Costco partnership for mail and specialty pharmacy, and page 11 prices mail order as discounts off average wholesale price with no dispensing fee. Nothing invoices claims at any related pharmacy at net acquisition cost plus a stated fee, and there is no documentation default. A no spread promise is a different protection from acquisition cost pricing, so the subject is not addressed.
No verbatim quote is available for this clause.
Model contract says
Claims dispensed by PBM-owned or economically related pharmacies (mail, specialty, retail) are invoiced at net acquisition cost plus a stated dispensing fee, with no retained margin and a documentation default if records are not produced.
Owned pharmacies are where PBMs earn the most margin. Pricing them at cost removes the incentive to steer patients into the PBM's own channels.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol partial, Claude Opus 5 missing, Gemini 3.1 Pro missing
Benchmark ceiling and cheapest lawful option
Weight 3 of 100, model section Section 3.3
Not addressed
All five models reached this verdict. The pricing summary on pages 10 and 11 sets discounts off average wholesale price but never caps what the plan pays against a published benchmark or a cash price, and there is no claim by claim test. The subject does not appear in the memoranda or the consultant notes on page 17. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
Plan claim cost can never exceed a published benchmark, for example the transparent cash price at a benchmark site, in any channel and for any drug, and the ceiling is checked claim by claim rather than on average.
A ceiling protects the plan from paying more through the contract than the public can pay in cash, and a claim level test stops a favorable average from hiding individual overcharges.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Cash price protection and deductible credit
Weight 3 of 100, model section Section 3.4
Not addressed
All five models reached this verdict. Nothing protects a member from paying more through the plan than the pharmacy cash price, and nothing addresses whether a cash purchase counts toward the deductible or out of pocket maximum. The closest passage is the point of sale rebate design on page 17, which lowers what a member pays at the counter but is not a cash price ceiling and says nothing about accumulators. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
A member never pays more than the pharmacy cash price or the cheapest lawful price, every cash purchase counts toward the deductible and out of pocket maximum, and no accumulator adjustment is applied against member assistance unless the sponsor elects it in writing.
Members routinely pay more through the plan than the cash price on the shelf, and a cash purchase that does not count toward the deductible punishes the member for finding the cheaper price.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
MAC list governance
Weight 5 of 100, model section Section 3.5
Not addressed
All five models reached this verdict. Maximum allowable cost appears on pages 10 and 11 only as the label for the current OptumRx generic pricing, while the proposed Envision generic rates are stated as discounts off average wholesale price. There is no statement that one list governs both what the county pays and what pharmacies are paid, no update schedule and no pharmacy appeal path. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
A single MAC list applies to both what the plan pays and what pharmacies are paid, it is disclosed, updated on a schedule, and pharmacies have an appeals path with defined timelines.
Two MAC lists (one for the plan, one for pharmacies) is the classic spread pricing mechanism.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
No spread pricing
Weight 9 of 100, model section Sections 2.5, 3.1, 3.5
Meets the model
Three of five models read this as good; Claude Opus 5 and Grok 4.6 called it partial because the sentence is a county summary of a proposal rather than executed adjudication language, and because the same pages present guaranteed discounts off average wholesale price that could still differ from what a pharmacy is paid. The pharmacy memorandum on page 10 states that Envision agreed to a complete pass through of all pharmacy discounts with no spread pricing on retail, mail or specialty pharmacy, page 11 shows the PBM's compensation as a separate $3.50 per employee per month administrative fee, and page 17 records that Envision will accept only that fee. That is the structure the model contract published by Mark Cuban and collaborators asks for, so the majority verdict stands. Confidence is held down because the enforceable clause, the definition of the pharmacy payment and any reconciliation duty are in the executed agreement, which is not in the file.
has agreed to a complete pass-
through of all pharmacy discounts, which includes no spread pricing on retail, mail or specialty pharmacy.
Page 10Model contract says
The plan pays exactly what the pharmacy is paid plus a disclosed administrative fee. The PBM retains no difference between the two, in any channel.
Spread pricing is an undisclosed markup on every claim. Eliminating it is the single clearest test of a transparent contract.
Model votes: Kimi K3 good, Grok 4.6 partial, GPT 5.6 Sol good, Claude Opus 5 partial, Gemini 3.1 Pro good
GPO and purchasing entity pass-through
Weight 4 of 100, model section Section 5
Not addressed
Four of five models read this as missing; GPT 5.6 Sol called it partial because the phrase all monies received or earned on page 10 is broad enough to reach purchasing entity revenue. No page names a group purchasing organization, rebate aggregator, volume credit or supplier payment, and there is no allocation method. Page 10 says Envision contracts directly with pharmacies rather than using rental networks, which speaks to network fees, not purchasing entity revenue. A general phrase without a clause is not a GPO pass through, so the subject is not addressed.
No verbatim quote is available for this clause.
Model contract says
Purchase discounts, volume credits and supplier payments received by the PBM or any related purchasing entity (including offshore GPOs) flow to the plan, allocated on disclosed drivers.
PBMs moved much of their manufacturer revenue into affiliated GPOs after rebate scrutiny increased. A contract that only covers rebates misses this.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol partial, Claude Opus 5 missing, Gemini 3.1 Pro missing
Plan sponsor controls the formulary
Weight 4 of 100, model section Section 6
Not addressed
Four of five models read this as missing; GPT 5.6 Sol called it partial because page 10 describes formulary management responsive to county needs. That sentence praises Envision's flexibility through formulary management, which is a vendor capability, not a grant of approval rights to the county. Nothing says the county approves the formulary or its changes, receives lowest net cost analysis or gets utilization management outcomes by drug. Page 9 lists a covered drug list as a bid evaluation criterion only. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
The plan sponsor approves the formulary and every change, receives lowest-net-cost analysis, and gets utilization management outcomes by drug each quarter.
A PBM-controlled formulary can favor high-list-price, high-rebate drugs that cost the plan more overall.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol partial, Claude Opus 5 missing, Gemini 3.1 Pro missing
Lowest net cost standard and book of business comparison
Weight 3 of 100, model section Section 6
Not addressed
All five models reached this verdict. There is no lowest net cost standard for formulary decisions and no annual comparison against Envision's whole book of business anywhere in the 18 pages. The claim repricing described on pages 10 and 17 was a one time procurement exercise, not an ongoing contractual standard, and nothing requires disclosure when a drug is placed above a cheaper equivalent. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
Formulary decisions must meet a lowest net cost standard for this plan, the PBM delivers an annual comparison against its whole book of business, and any drug placed above a cheaper equivalent is disclosed with the reason.
Rebate driven formularies can favor a high list price drug that returns more rebate but costs the plan more after the rebate. A stated standard and a book of business comparison make that trade visible.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Pharmacy network protections and anti-steering
Weight 7 of 100, model section Section 7
Weaker than the model
Four of five models read this as partial; Gemini 3.1 Pro called it missing because the executed network terms are absent. Page 10 says members may fill 90 day prescriptions at retail instead of being required to use mail order, and pages 10 and 11 price a 90 day retail network, which is a real protection against steering members into one channel. Page 10 also describes more than 67,000 retail pharmacies contracted directly rather than rented. None of the other model protections appear, including finality of adjudicated claims, a bar on retroactive clawbacks, a ban on volume conditioned reimbursement and an any willing pharmacy right, and nothing forbids steering to the Costco mail and specialty partner.
They also have the option for health plan members to fill 90-day prescriptions at retail locations instead of requiring the use of the mail-order service.
Page 10Model contract says
Claims are final when adjudicated (no retroactive clawbacks), reimbursement is not conditioned on volume, the PBM may not steer claims into its own pharmacies, and any willing pharmacy may participate at the plan's terms.
Steering and clawbacks push independent pharmacies out and route patients to PBM-owned channels, reducing choice and often raising cost.
Model votes: Kimi K3 partial, Grok 4.6 partial, GPT 5.6 Sol partial, Claude Opus 5 partial, Gemini 3.1 Pro missing
Most favored pricing and market check
Weight 3 of 100, model section Section 8
Not addressed
All five models reached this verdict. Pricing is fixed in the summary tables on pages 10 and 11 with no most favored client promise and no annual market check, and there is no meet, credit or release remedy. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
The plan automatically receives pricing at least as good as any comparable client, and can run an annual market check with a meet, credit or release remedy.
Multi-year contracts drift out of market. Without a market check the plan is locked into stale pricing.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Audit rights, data access and reporting
Weight 6 of 100, model section Section 9
Not addressed
Four of five models read this as missing; Kimi K3 called it partial because page 9 lists claim level repricing and standard, custom and ad hoc reporting among the bid evaluation criteria. Those lines describe how proposals were scored, not a continuing right for the county to audit with its own auditor, receive claims level financial data on a schedule or see claims before adjudication. Page 10 describes integrating pharmacy and medical data for case management, which is a service, not an audit right. No audit article appears in the packet, so the subject is not addressed.
No verbatim quote is available for this clause.
Model contract says
The plan may audit with its own auditor, receives full claims-level and financial data on a schedule without asking, has pre-adjudication claims access, and no data is withheld as proprietary.
You cannot enforce what you cannot see. Audit and data rights are how every other clause gets verified.
Model votes: Kimi K3 partial, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Plan sponsor owns all plan data
Weight 4 of 100, model section Sections 9.4, 9.5
Not addressed
All five models reached this verdict. The only data language is on page 10, where Envision and United Healthcare commit to integrating pharmacy data with medical data for case and disease management, which describes a use of data rather than who owns it. Nothing makes the county the owner of claims, eligibility, accumulator or rebate data, requires machine readable delivery or bars reuse or sale. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
The plan sponsor is the sole owner of claims, eligibility, accumulator, rebate and every derived data set, the PBM delivers it in machine readable form on request, and the PBM may not reuse, license or sell it.
Data ownership decides who can switch vendors, run an audit or check a guarantee. A PBM that owns or licenses plan data can charge for it, withhold it at termination, or resell it.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Performance guarantees with real remedies
Weight 2 of 100, model section Section 10
Weaker than the model
All five models reached this verdict. Pages 2 and 10 record substantial performance guarantee monies for service and a proposed 9.5 percent spend reduction guarantee over current pharmacy costs, and page 11 adds rebate minimums of $17.35 per Brand Rx at retail and $66.60 per Brand Rx at mail. Guarantees plainly exist, but the packet does not say how they are measured, whether they use Travis County data, whether shortfalls are paid in cash or as capped credits, or how often they are reconciled. The guarantee exhibit that would answer those questions is not part of the backup.
Envision included substantial performance guarantee monies for the performance and service of the plan. In addition to service guarantees, Envision is proposing a 9.5% spend reduction guarantee over the current pharmacy costs.
Page 10Model contract says
Guarantees are measured quarterly on plan-specific data, reconciled in cash, and shortfalls are paid dollar for dollar rather than as capped credits.
Guarantees that are capped or measured on book-of-business data rarely pay out.
Model votes: Kimi K3 partial, Grok 4.6 partial, GPT 5.6 Sol partial, Claude Opus 5 partial, Gemini 3.1 Pro partial
Self executing enforcement and two strikes exit
Weight 3 of 100, model section Sections 10, 12.5
Not addressed
All five models reached this verdict. Nothing makes a missed guarantee or a late report trigger liquidated damages or an automatic credit, there is no two strikes exit and no regulatory integrity trigger. Pages 2 and 10 say guarantee money was proposed but do not say how it is enforced. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
Missed guarantees and late reports trigger liquidated damages or automatic credits without the plan having to prove damages, and a second material breach or regulatory integrity failure lets the sponsor terminate without penalty.
A right the plan has to litigate to enforce is rarely enforced. Automatic remedies and a defined exit after a repeat failure change the PBM incentives without a lawsuit.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Amendments only in writing
Weight 2 of 100, model section Section 11
Not addressed
All five models reached this verdict. Page 3 labels the contract type as annual and unilateral and page 16 says the three year term runs unless changed by Travis County, but neither line is an amendment procedure. Nothing requires amendments to be signed by both parties or bars changes to economic terms by notice or updated exhibit. The executed agreement is not part of this agenda backup and nothing in the 18 pages is redacted, so the subject is absent rather than unreadable.
No verbatim quote is available for this clause.
Model contract says
No amendment is effective unless signed by both parties; the PBM cannot change economic terms by notice, portal update or updated exhibit.
Unilateral amendment rights let a PBM quietly rewrite pricing mid-term.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 missing, Gemini 3.1 Pro missing
Termination without penalty
Weight 2 of 100, model section Section 12
Not addressed
Four of five models read this as missing; GPT 5.6 Sol called it partial because the page 16 phrase unless changed by Travis County suggests county control over the term. Page 3 gives a contract period of October 1, 2014 through September 30, 2015 and page 16 gives a three year term ending on a September 31, 2017 date that does not exist, so the packet does not even agree with itself on the term. Neither passage grants termination for convenience, sets notice, waives penalties or protects earned rebates and guarantee amounts on exit. A statement of term is not a termination clause, so the subject is not addressed.
No verbatim quote is available for this clause.
Model contract says
The plan may terminate for convenience with reasonable notice and without penalty, earned amounts are never forfeited, and a regulatory integrity event is grounds for termination.
Termination penalties and forfeited rebates are how plans get trapped in bad contracts.
Model votes: Kimi K3 missing, Grok 4.6 missing, GPT 5.6 Sol partial, Claude Opus 5 missing, Gemini 3.1 Pro missing
Specialty drug pricing controls
Weight 3 of 100, model section Sections 1.16, 3.2(d)
Weaker than the model
Three of five models read this as partial; Gemini 3.1 Pro and Kimi K3 called it missing because there is no specialty definition, rate table or ceiling. Page 10 extends the no spread promise to specialty pharmacy and names a Costco partnership for mail and specialty, which is the only specialty control a reader can see. The rate tables on pages 10 and 11 cover retail, 90 day retail and mail order and contain no specialty discounts, dispensing fees or drug level ceilings, and nothing prices specialty claims at acquisition cost. The subject is touched but far short of the model, and the executed specialty terms are not in the file.
Envision’s proposal includes a partnership with Costco for mail and specialty pharmacy and has agreed to a complete pass-
through of all pharmacy discounts, which includes no spread pricing on retail, mail or specialty pharmacy.
Page 10Model contract says
Specialty is defined by function, drug-level rates act as a ceiling in every channel, and specialty claims at affiliate pharmacies are priced at acquisition cost.
Specialty drugs are a minority of claims but the majority of spend, and most are dispensed by PBM-owned specialty pharmacies.
Model votes: Kimi K3 missing, Grok 4.6 partial, GPT 5.6 Sol partial, Claude Opus 5 partial, Gemini 3.1 Pro missing
Gag clause (red flag)
Weight 0 of 100, model section Sections 9, 14
Not addressed
Three of five models read this as missing; Claude Opus 5 and Kimi K3 called it good because nothing in the 18 pages treats pricing, rebates or fees as confidential and the discount rates, dispensing fees, rebate minimums and administrative fee are printed in a public agenda record. The packet contains no confidentiality article and no pharmacy communication clause, because it is not the executed agreement, so whether the contract restricts the county from sharing pricing data or restricts pharmacies from telling members about cheaper options cannot be read from this file. Nothing is redacted, so the verdict is missing rather than unclear. A missing verdict does not raise the gag flag, and readers should not take that as proof the signed contract has no confidentiality clause.
No verbatim quote is available for this clause.
Model contract says
Nothing in the contract stops the plan from sharing pricing, rebate or performance data with its own advisors, auditors, members, the public or lawmakers, and nothing stops pharmacies from telling patients about cheaper options.
Confidentiality clauses that treat pricing terms as trade secrets are what keep taxpayers from seeing what their government pays. This is why the A&M contract is partly blacked out.
Model votes: Kimi K3 good, Grok 4.6 missing, GPT 5.6 Sol missing, Claude Opus 5 good, Gemini 3.1 Pro missing
The council
Five models, every verdict published
The full independent review from each model, the agreement grid, and the synthesis that reconciles them.
Five models from five providers read the full extracted text independently, without seeing one another, and scored every rubric item. Unanimous verdicts stand. Where the council split, the reconciled verdict and the reasoning behind it are written out in the clause by clause section above, and the split is marked here so nobody has to take the resolution on faith. Items a model could not read because of redaction are unclear and excluded from the score. The council agreed outright on 12 of 23 clauses.
Swipe the grid sideways to see every seat and the published verdict.
Clause by clause verdicts by each council model and the published consensus
Clause
Claude Opus 5
Gemini 3.1 Pro
GPT 5.6 Sol
Grok 4.6
Kimi K3
Published
Fiduciary duty and loyalty
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Full disclosure of compensation and conflictsSplit
Weaker than the model
Not addressed
Weaker than the model
Not addressed
Weaker than the model
Weaker than the model
Ownership and affiliate disclosure exhibit
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Definitions that close loopholes
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
100% pass-through of manufacturer revenueSplit
Weaker than the model
Weaker than the model
Weaker than the model
Weaker than the model
Meets the model
Weaker than the model
Affiliate pharmacy pricing at acquisition costSplit
Not addressed
Not addressed
Weaker than the model
Not addressed
Not addressed
Not addressed
Benchmark ceiling and cheapest lawful option
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Cash price protection and deductible credit
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
MAC list governance
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
No spread pricingSplit
Weaker than the model
Meets the model
Meets the model
Weaker than the model
Meets the model
Meets the model
GPO and purchasing entity pass-throughSplit
Not addressed
Not addressed
Weaker than the model
Not addressed
Not addressed
Not addressed
Plan sponsor controls the formularySplit
Not addressed
Not addressed
Weaker than the model
Not addressed
Not addressed
Not addressed
Lowest net cost standard and book of business comparison
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Pharmacy network protections and anti-steeringSplit
Weaker than the model
Not addressed
Weaker than the model
Weaker than the model
Weaker than the model
Weaker than the model
Most favored pricing and market check
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Audit rights, data access and reportingSplit
Not addressed
Not addressed
Not addressed
Not addressed
Weaker than the model
Not addressed
Plan sponsor owns all plan data
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Performance guarantees with real remedies
Weaker than the model
Weaker than the model
Weaker than the model
Weaker than the model
Weaker than the model
Weaker than the model
Self executing enforcement and two strikes exit
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Amendments only in writing
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Not addressed
Termination without penaltySplit
Not addressed
Not addressed
Weaker than the model
Not addressed
Not addressed
Not addressed
Specialty drug pricing controlsSplit
Weaker than the model
Not addressed
Weaker than the model
Weaker than the model
Not addressed
Weaker than the model
Gag clause (red flag)Split
Meets the model
Not addressed
Not addressed
Not addressed
Meets the model
Not addressed
Council synthesis: where the five models agreed and where they split
How the council worked
Five frontier models read the same 18 page text: the Travis County Commissioners Court backup for Item 21 of the June 10, 2014 agenda, awarding pharmacy benefit management services to Envision Pharmaceutical Services, LLC from October 1, 2014 (page 1). The models were Claude Opus 5 (Anthropic), GPT 5.6 Sol (OpenAI), Gemini 3.1 Pro (Google), Grok 4.6 (xAI) and Kimi K3 (Moonshot AI). Each read independently against the same 23 item rubric, derived from the open source model PBM contract published by Mark Cuban and collaborators. We reconciled the reads under a fixed rule: a verdict held by three or more seats stands; with no majority the middle vote is taken after re-reading the cited pages; unclear is reserved for redacted text. Nothing here is redacted, so where the backup lacks contract language the verdict is missing. Every seat's vote on every item is published on this page.
One fact shaped every read. This is an award packet, not the executed agreement. It holds staff memoranda, a pricing summary (pages 10 and 11) and the consultant's recommendation (pages 16 to 18). The signed contract is not in the file.
Where the five models agreed
No governance language. All five found no fiduciary duty, no ownership or affiliate exhibit, no definitions article, no amendment rule, no self executing enforcement and no data ownership clause.
No price ceilings. All five found no benchmark or cash price ceiling, no cash price protection for members, no MAC list governance, no most favored pricing and no lowest net cost standard. MAC appears on page 11 only as the label for the outgoing OptumRx arrangement.
Guarantees exist but their mechanics do not. All five voted partial on performance guarantees: pages 2 and 10 record substantial performance guarantee monies and a proposed 9.5% spend reduction guarantee, and page 11 adds rebate minimums of $17.35 per Brand Rx at retail and $66.60 per Brand Rx at mail, but nothing says how any guarantee is measured, reconciled or paid.
Where they split, and how it was resolved
Eleven items split. The majority decided each one; none was overruled.
No spread pricing. GPT 5.6 Sol, Gemini 3.1 Pro and Kimi K3 voted good on the page 10 statement that Envision agreed to a complete pass through of all pharmacy discounts with no spread pricing on retail, mail or specialty pharmacy, read with the $3.50 per employee per month administrative fee (page 11). Claude Opus 5 and Grok 4.6 voted partial because the sentence summarizes a proposal, not executed adjudication language. Majority good, confidence 0.6.
Pass through of manufacturer revenue. Four voted partial; Kimi K3 voted good on the rebate minimums. The page 10 promise to pass through 100% of all monies received or earned is broad, but there is no definition of the money covered, no payment schedule and no exclusion list.
Disclosure of compensation. Claude Opus 5, GPT 5.6 Sol and Kimi K3 voted partial on the page 17 fee only promise; Gemini 3.1 Pro and Grok 4.6 voted missing because one fee promise is not a disclosure covenant.
Specialty controls. Three voted partial because the no spread sentence reaches specialty and names a Costco partnership (page 10); Gemini 3.1 Pro and Kimi K3 voted missing because there is no specialty definition or rate table.
Network protections. Four voted partial on the page 10 option to fill 90 day prescriptions at retail instead of mail order; Gemini 3.1 Pro voted missing.
Gag clause. GPT 5.6 Sol, Gemini 3.1 Pro and Grok 4.6 voted missing because there is no confidentiality article to read either way; Claude Opus 5 and Kimi K3 voted good because every price in the packet is printed in a public record. Majority missing, which leaves the gag flag unset without proving the signed contract is clean.
Five lone dissents. GPT 5.6 Sol alone voted partial on affiliate pricing, GPO pass through, formulary control and termination, reading the broad pass through phrase and the page 16 words unless changed by Travis County generously. Kimi K3 alone voted partial on audit rights because page 9 lists claim level repricing as a bid criterion. Each fell to a four vote missing majority: a proposal summary and an evaluation criterion are not contract rights.
What only one model caught
Kimi K3 flagged the $390.00 per case fee for physician reviewed prior authorizations on page 11, although the column layout places that figure under the current OptumRx arrangement rather than the Envision proposal, and the consultant's admission on page 17 that no TPA proposer repriced medical claims to the claim level, so all their numbers were thrown out.
Claude Opus 5 noticed that the mail order rebate minimum on page 11 sits in a row labeled Retail Rebate Minimum, an apparent labeling error.
Grok 4.6 quoted the page 10 warning that additional fees for services are included in the Contract, so the fee table is incomplete.
None of the five noted that the mail order brand discount moved from AWP-22% to AWP-20.40% (page 11), a smaller discount than the county had before.
The bottom line
The reconciled score is 23.5 out of 100, a Red flag grade, with a scored weight of 100 and an unclear weight of 0 because nothing is redacted. The gag flag is not raised. Sixty two of the 100 weight points went to items the packet does not address. The five seats' own scores ran from 15.5 (Gemini 3.1 Pro) and 16 (Grok 4.6) through 19 (Claude Opus 5) to 30.5 (Kimi K3) and 31 (GPT 5.6 Sol), a spread driven by how much credit each gave to promises described in memoranda rather than shown in signed text. The number describes what Travis County disclosed to the public in June 2014, not the whole deal.
The same contract written up for different readers: the people who negotiate these agreements and the people who pay for them.
Good news, bad news: the short version
Good news
No spread pricing, in writing, in every channel. The county's pharmacy memorandum says Envision agreed to a complete pass through of all pharmacy discounts, which includes no spread pricing on retail, mail or specialty pharmacy (page 10). That is the clearest test of a transparent PBM deal.
One fee, and it is printed. Envision's compensation is an administrative fee of $3.50 per employee per month (page 11), and the consultant records that Envision will put in writing that it will accept only that fee and no other money tied to the county's drug spend (page 17).
100% pass through of manufacturer money. Staff say Envision agreed to pass through 100% of all monies received or earned through the utilization of Travis County and its members (page 10), with rebate floors of $17.35 per Brand Rx at retail and $66.60 per Brand Rx at mail (page 11).
Rebates at the counter. Envision's design delivers rebates at the point of sale rather than after a year end reconciliation (page 17).
Members are not forced into mail order. Members may fill 90 day prescriptions at retail instead of being required to use mail order (page 10) (page 10).
Guarantees with money behind them. The proposal includes substantial performance guarantee monies and a 9.5% spend reduction guarantee over current pharmacy costs (page 10).
Bad news
The contract is not in the packet. Every promise above is a staff or consultant description of a proposal. Definitions, audit rights, data ownership, termination and confidentiality all live in a signed agreement that was not attached.
Nothing defines the money. Rebate, pharmacy discount and specialty are used on pages 10 and 11 without definitions, so a reader cannot tell whether the 100% pass through reaches administrative fees, data fees or other manufacturer payments.
No audit right, no data ownership. Reporting and claim level repricing appear only as bid scoring criteria (page 9). Nothing grants an audit right or makes the county the owner of its claims data.
No ceiling on what the plan pays. Pricing is discounts off average wholesale price (pages 10 and 11) with no cash price or benchmark ceiling, no MAC list governance and no market check.
The mail order brand discount got weaker. Mail order brand pricing moved from AWP-22% today to AWP-20.40% under the proposal (page 11).
Fees outside the table. The memorandum warns that additional fees for services are included in the Contract (page 10), and the visible list already includes a case review charge of $8.00 per case (page 11).
The term does not agree with itself. Page 3 lists a contract period of October 1, 2014 through September 30, 2015; page 16 lists a three year term ending September 31, 2017, a date that does not exist.
What we could not see
Nothing in the 18 pages is redacted, so no item was excluded from the score as unclear. What is absent is the executed agreement itself, with its definitions, rate exhibits, guarantee schedule, audit and confidentiality articles. Sixty two of the 100 weight points went to items the packet does not address, and those scored zero.
The number
The reconciled score is 23.5 out of 100, a Red flag grade, on a scored weight of 100 with no unclear weight; the gag flag is not raised because there is no confidentiality clause in the file to read either way. The score describes what Travis County disclosed in a 2014 award packet, not the whole deal. What would move it most is the executed agreement: if the signed text turned the page 10 pass through statement into a defined, auditable clause with a payment schedule, that item alone (weight 11) would add up to 5.5 points, and real audit and data ownership articles (weights 6 and 4) would add up to 10 more.
Check our work
Read a page
Pull the text of any page we extracted and compare it with the original.
We already have this document and the posted copy includes its pricing attachments, so you do not need to request the contract. What a request can still add is the incorporated proposal, the claims and rebate reports the plan receives, and any later amendments. Generate a Texas Public Information Act letter for Travis County that names those pieces, send it in your own name, and log what comes back.
The model contract we score against is the open source PBM contract published by Mark Cuban and collaborators, version 3.5. BetterBuy Rx did not write it and is not affiliated with its authors. Read the original on LinkedIn.