Which harness runs Jev?

This comparison uses Browser Use’s dedicated Jev Ultrafast harness. Jev is a decision model accessed through OpenRouter’s Decisions API. A general-purpose chat harness or LangChain classifier integration is not a drop-in browser harness for it.

Are model names hidden?

No. This is a named comparison with visible models and costs. You can vote for either result or a tie; your choice stays outside the blind community rankings.

Can I change the options?

No. This free combo fixes Jev Ultrafast with Jev 1.13 against Codex with GPT-5.6 Sol. Other agent and model combinations run from the other comparison pages at normal usage pricing.

The free combo covers Jev, DeepSeek V4.1 Flash, Codex, browser usage and compute for up to 24 hours, ending when you vote. Continued use is billed normally.

Jev Ultrafast vs Codex

Two live browsers, one search. Watch both, inspect the results, then choose your favorite.

Leaderboard

Jev UltrafastJev 1.13FreeDeepSeek V4.1 Flash handles text entry

CodexGPT-5.6 SolMediumFree

Search and filter

Start with a flight search.

Both agents start on Google Flights with the same browser size. The task stops at matching results, before any booking.

Task verification is shown separately from the agent’s own completion claim. A blocked page or incomplete search stays visible.

Watch live
See both browser views together and expand either one. Viewing does not control the agent’s browser.
Inspect the work
Compare current pages, actions, elapsed time and the matching search settings.
Compare metered costs
Jev chooses browser actions; DeepSeek V4.1 Flash supplies text-entry values. The free combo covers both models and browser usage until you vote or the free window ends.

Which option do users prefer?

2 tries · 0 votes · Same agents and models

No attributable votes for these agents and models yet. Try a real task and choose the result you prefer.

Recent user choices

See which results other people preferred. Their tasks and files stay private.

No public comparison records to show.

Model benchmarks

Artificial Analysis

Intelligence v4.3

General capability across the lab's whole suite.

Not published

47.1

Jev UltrafastJev 1.13

CodexGPT-5.6 Sol (max)

Coding

The coding subset of the same suite.

Not published

77.4

Jev UltrafastJev 1.13

CodexGPT-5.6 Sol (max)

Agentic

Tasks the model runs in several steps.

Not published

50.5

Jev UltrafastJev 1.13

CodexGPT-5.6 Sol (max)

These describe the model on its own, not the harness, the tools or the whole setup this page compares — and the lab runs each model at maximum effort, which is not what the composer above is set to. Scores are only comparable within one index version.

Source: Artificial Analysis — their published model index, taken 2026-09-16.

Agent performance on real tasks

Published Terminal-Bench 4.0 runs. Each result names its tested configuration.

Accuracy · Terminal-Bench 4.0

Percent of trials the whole configuration solved.

Not published

37.3%

Jev UltrafastJev Ultrafast · Jev 1.13

CodexCodex · GPT-5.6 Sol123 of 330 trials · max effort

The whisker is the 95% confidence interval. Two configurations whose whiskers overlap are not separated by this benchmark, however far apart their bars look.

Source: Terminal-Bench — their published runs of a harness and model together, taken 2026-09-17.