@p0 (Parallel Web Systems) says on a generic decisions benchmark OpenAI, Jev, and Parallel all score about 90%, but on live-web questions OpenAI drops 2.2× and Jev drops 2.6×. They built PWS-Decisions so generic benches cannot hide that gap, and link the benchmark blog.

Parallel Web Systems @p0

2026-11-05EN

835 views
14 likes
4 bookmarks
1 reposts

On a generic decisions benchmark, OpenAI, Jev, and Parallel all score about 90%. On questions that depend on the live web, OpenAI drops 2.2x, and Jev drops 2.6x…

About this case

Benchmark post. Parallel Web Systems (@p0) reports ~90% for OpenAI, Jev, and Parallel on a generic decisions bench, then 2.2× and 2.6× drops for OpenAI and Jev on live-web questions. They introduce PWS-Decisions and link https://parallel.ai/blog/pws-decisions-benchmark with a chart image.

Original posthttps://x.com/p0/status/2118456575761473863

Related cases

@sardine says Jev caught 27 of 29 fraudulent transactions in a September fraud ring, the most of any approach they tested, and that Jev is now a node in Sardine Flow.@sardine
@noahsark769 had @typesafeai Jev look at and categorize all ads for the top 680 brands on @replohq's industry landscape list. About 67% of these brands have some ads with customer quotes, and of those brands about 4–10% of the ads they're running have customer quotes.@noahsark769
@Centific_CAIR reports Jev matched Claude Sonnet 4 spotting harmful AI replies (F1 0.758 vs 0.717) at 176× lower cost on 498 samples; newer Claude models often refused to judge.@Centific_CAIR
@Mav1972 checked 54 storefront screenshots with short yes/no questions. Jev answered in 0.4 s a call, caught 19 of 22 planted defects, and raised 0 of 24 false alarms on harmless changes.@Mav1972