The measurements disagree, and that is the finding: of 12 runs, 6 runs place Claude Opus 5's prisma boundary inside 7.x (7.0.0 (2025-11-19)) and 6 runs below it (6.7.0 (2025-04-29), 6.8.0 (2025-05-15), 6.10.0 (2025-06-17) and 6.15.0 (2025-08-27)). A boundary is a draw, not a constant. That is ~5 to ~13 months below the training cutoff the subject stated in those runs (2026-05). 16 findings are currently charged against Claude Opus 5 on prisma (3x S1 breaks-build, 10x S2 silently-wrong, 1x S3 deprecated, 2x S4 wrong-metadata), each reproduced in a published run and checked against prisma's own release notes. Read the boundary precisely: it is the newest release whose contents the model can correctly attribute to that release, not the newest prisma feature it can use. Past it a model often writes working code with a newer API while naming the wrong release for it.
Answer class split, decided by one rule applied to every page
of this kind: the measured boundaries fall on both sides of the first release of the major named in the question. Every figure below is read from
the dataset at build time; nothing on this page is written by hand.
| Battery | Newest release it can place | Oldest it cannot | Lag vs stated cutoff |
|---|---|---|---|
| v1 2026-08-31 | 7.0.0 2025-11-19 |
7.1.0 2025-12-03 |
~5 months |
| v1r-a 2026-09-01 | 7.0.0 2025-11-19 |
7.1.0 2025-12-03 |
~5 months |
| v1r-b 2026-09-01 | 6.7.0 2025-04-29 |
6.9.0 2025-06-03 |
~12 months |
| v2-a 2026-09-05 | 6.7.0 2025-04-29 |
6.8.0 2025-05-15 |
~12 months |
| v2-b 2026-09-05 | 7.0.0 2025-11-19 |
7.1.0 2025-12-03 |
~5 months |
| v3-a 2026-09-05 | 6.8.0 2025-05-15 |
6.9.0 2025-06-03 |
~12 months |
| v3-b 2026-09-05 | 6.10.0 2025-06-17 |
6.11.0 2025-07-01 |
~11 months |
| v3-c 2026-09-05 | 7.0.0 2025-11-19 |
7.1.0 2025-12-03 |
~6 months |
| v3-d 2026-09-05 | 6.15.0 2025-08-27 |
6.16.0 2025-09-10 |
~9 months |
| v3-e 2026-09-05 | 7.0.0 2025-11-19 |
7.1.0 2025-12-03 |
~6 months |
| v5-c 2026-09-06 | 6.7.0 2025-04-29 |
6.8.0 2025-05-15 |
~13 months |
| v5-d 2026-09-06 | 7.0.0 2025-11-19 |
7.1.0 2025-12-03 |
~6 months |
2 further runs of this pair established only one end of the interval, or measured the instrument rather than the library; they are listed under Runs below.
12 measurements of this pair, giving 5 different boundaries — 204 days apart, 13 releases apart. Of those, 2 came from the same stored prompt file, sent concurrently and blind: they disagreed by 204 days, 13 releases apart.
prisma published 21 minor or major releases in the twelve months before this model’s stated cutoff, which is the scale a spread should be read against.
“Chargeable” means the change was published before this model’s own stated cutoff, so it had the opportunity to know it.
| Severity | Belief | Changed in | Chargeable | Proof |
|---|---|---|---|---|
| S1breaks-build | prisma.config.ts written with an adapter key, removed in the same release that made the file mandatoryprisma.config.ts adapter key |
7.0.0 2025-11-19 |
yes | run · source |
| S1breaks-build | SQLite adapter imported under its pre-7.0.0 class namePrismaBetterSQLite3 |
7.0.0 2025-11-19 |
yes | run · source |
| S1breaks-build | migrate diff invoked with both removed flag familiesprisma migrate diff |
7.0.0 2025-11-19 |
yes | run · source |
| S2silently-wrong | Denies that Prisma ships a first-party SQLCommenter plugin@prisma/sqlcommenter-query-insights |
7.1.0 2025-12-04 |
yes | run · source |
| S2silently-wrong | Denies that db push and migrate dev accept a --url flagprisma db push --url |
7.2.0 2025-12-17 |
yes | run · source |
| S2silently-wrong | Denies that the generator has a size/speed optioncompilerBuild |
7.3.0 2026-01-21 |
yes | run · source |
| S2silently-wrong | Denies that a Prisma schema can express a partial index, and writes the denial into the schema file as an instruction not to add the constraint@@index([...], where: ...) / @@unique([...], where: ...) |
7.4.0 2026-02-11 |
yes | run · source |
| S2silently-wrong | Recites the current deny list and then puts $transaction back into it, denying nested transactions and rejecting the working calltx.$transaction() |
7.5.0 2026-03-11 |
yes | run · source |
| S2silently-wrong | Denies the Prisma CLI has any command for attaching a project to an existing Prisma Postgres databaseprisma postgres link |
7.6.0 2026-03-27 |
yes | run · source |
| S2silently-wrong | Denies the pg adapter constructor takes a bare connection string, and rewrites the working callnew PrismaPg(connectionString) |
7.6.0 2026-03-27 |
yes | run · source |
| S2silently-wrong | Denies @prisma/adapter-pg exposes any control over prepared-statement names, and recommends running the query outside Prisma insteadstatementNameGenerator |
7.6.0 2026-03-27 |
yes | run · source |
| S2silently-wrong | Denies the CLI has a single command that sequences the whole project setupprisma bootstrap |
7.7.0 2026-04-07 |
yes | run · source |
| S2silently-wrong | Denies the query plan cache can be sized or disabled from the PrismaClient constructor, and rejects the working line that does itqueryPlanCacheMaxSize |
7.8.0 2026-04-22 |
yes | run · source |
| S3deprecated | Connection string left in the schema's datasource block rather than the config filedatasource url |
7.0.0 2025-11-19 |
yes | run · source |
| S4wrong-metadata | Set a MongoDB project up on the version that dropped MongoDBMongoDB support |
7.0.0 2025-11-19 |
yes | run · source |
| S4wrong-metadata | Version knowledge stops at 7.0.0, five months before its own stated cutoff | 7.1.0 2025-12-03 |
yes | run · source |
The prisma correction pack states what is true now for each corrected fact, with a primary-source citation, as markdown you can paste into a rules file (raw). What a correction pack measurably changed when one was tested — a pre-registered run on zod — is on the benchmark page, including where it changed nothing.
All of them at once: Which Claude model knows prisma 7 best?
Other subjects on prisma:
Claude Opus 5 on the other libraries: