Published
Updated
Violoop ran the same desktop harness with Jev and with Opus 5.5 on one QuickBooks bill
This topic was created 31 days ago, and the information it contains may have evolved or changed since then.
Violoop posted four live desktop runs on September 29. On one prepared QuickBooks bill, the same harness made 123 decisions with Jev in 127.3 seconds and with Claude Opus 5.5 in 978.9 seconds. Estimated model cost was $0.0027 against $1.38. The Tetris snapshot used a faster gravity setting for Jev. We did not rerun the tasks.
Violoop posted on September 29 that its desktop harness can hand the next action to Jev. The harness narrows the current screen to a short list of valid actions. Jev picks one. The write-up, published the same day, calls the four runs live team demonstrations, and says they are not an independent benchmark.
The matched comparison is one prepared bill in QuickBooks Desktop. Both sides used the same harness and made 123 model decisions. All 30 rows were checked as they were entered. Claude Opus 5.5 took 978.9 seconds, with a median decision of 7,181 ms and an estimated model cost of $1.38. Jev took 127.3 seconds, median 390 ms, estimated cost $0.0027. The page’s “about 7.7 times” is that pair of clocks. 978.9 divided by 127.3 is 7.7. The median ratio is about 18. The cost ratio, $1.38 against $0.0027, is about 510, which the post rounds to about 500. Both cost cells are list rates for the model, not the rest of the machine.
Tetris is the same seed, read at one snapshot. Jev had placed 282 pieces and cleared 111 lines. Opus 5.5 had placed 71 and cleared 26. Average decision time was 519 ms against 2,331 ms, and the rounded model cost was $0.01 against $0.27. Jev was also on a faster gravity setting, 0.2 seconds against 0.4. The page records zero errors for either model on that seed, and says the snapshot is not a general reliability claim.
Two further runs have no second model. Jev ranked 450 Kickstarter comments, 279 top-level threads and 171 replies, in 30 seconds at about $0.014. It finished a 30-question form in 97.6 seconds, with all 30 answers verified, at about $0.002. Neither run prints an error count against a person or against another model.
The page says the team is researching whether a model like Jev can run on Violoop hardware, and that this is not a current local product. The post’s image is a title card for the same four tests. We did not rerun them.
This site's reading
Editorial notes evaluating claims against primary sources, contextualizing findings alongside related implementations, and defining technical terms.
Verify
Post by @violoop, September 29, 2026, 10:21 UTC. The attached image is a title card: up to 7.7 times faster and up to 500 times cheaper, four tests, and a line that more tasks are still to come. The write-up at violoop.ai, dated the same day, is the table.
QuickBooks Desktop, one prepared bill, same harness, 123 model decisions each: Opus 5.5 978.9 seconds, median decision 7,181 ms, estimated model cost $1.38, 30 of 30 rows checked during entry. Jev 127.3 seconds, median 390 ms, $0.0027, 30 of 30.
The page says the cost figures use model list rates and cover model usage only. Tetris, same seed, at the captured snapshot: Opus 71 pieces and 26 lines, average decision 2,331 ms, rounded model cost $0.27, gravity 0.4 seconds. Jev 282 pieces and 111 lines, 519 ms, $0.01, gravity 0.2 seconds.
Zero errors recorded for either model on that seed. Comment triage is Jev only: 450 comments, 279 top-level threads and 171 replies, 30 seconds, about $0.014, no accuracy. A 30-question form is Jev only: 97.6 seconds, 30 answers verified, about $0.002.
The byline on the page is BVIO marketing. We did not rerun the four tasks.
Compare
Browser Use's flights clip is one published Jev path at about 7 seconds, without an Opus row on the same harness. Stagehand is a different computer-use stack. Ashutosh Mathore's Tetris bake-off scores legal landings and points.
This Tetris snapshot changes the gravity between the two models, so the piece count is not a matched clock. The OCR computer-use note is one screenshot and a dollar figure, with no task success rate.
Terms
- 127.3 seconds
- Jev's total time on the one QuickBooks bill in the write-up. Opus 5.5 on the same harness and the same 123 decisions took 978.9 seconds. Both checked 30 of 30 rows during entry.
- $0.0027
- Estimated model cost for the Jev QuickBooks run, from model list rates. The Opus 5.5 cell is $1.38. The page says this is model usage only.
- gravity 0.2 s
- The Tetris setting on the Jev side of the snapshot. Opus 5.5 played at 0.4 seconds. Piece counts at the snapshot are 282 and 71.