DEV Community: BenchTested

ClickUp Review 2026: Free Plan Tested on Windows (8 Steps, 7.4/10)

BenchTested — Fri, 30 Oct 2026 12:27:02 +0000

This is a republished copy of our review. The original, with all 25 screenshots, the full pass/fail checklist and pricing, lives at benchtested.app/apps/clickup.

Verdict: 7.4/10, Recommended. Best for small teams that want lists, boards, Gantt charts, automations and AI in one free workspace and do not mind a busy, heavy app. Ranked #1 of 5 project management tools we have tested.

ClickUp passed every step of our project-management test on the free plan: all the views, a working automation, dependencies, recurring tasks, a 500-row import and an AI that got the facts right and asked before changing anything. It is heavy, though: the Windows app needs about 14 seconds and up to 2 GB of memory, and the import and setup make you do more by hand than they should.

How we tested

Windows 11 Pro in a clean virtual machine (4 vCPU, 8 GB RAM, 1920×1080), ClickUp desktop app 3.5.402 for Windows, a new account on Free Forever and a second test account for permissions. Eight steps, the same for every project management tool:

Step	Task	Result
P1	Build a 40-task product launch project from a written brief	Pass, ≈ 200 actions
P2	Show it as list, board, calendar and Gantt	Pass, ≈ 2 min
P3	Status automation, dependency, weekly recurring task	Pass, ≈ 7 min
P4	Import a 200-row CSV with tricky characters	Partial, ≈ 12 min
P5	Import 500 rows, open a 500-task board	Pass, import < 2 min
P6	Invite a second user, then restrict them	Pass, ≈ 5 min
P7	AI: summarise, write subtasks, "what is overdue?"	Pass, ≈ 1–1.5 min each
P8	Export everything, delete the workspace, uninstall	Partial, ≈ 7 min

What works

Every view we needed is free: list, board, calendar and Gantt, with no upgrade prompt
Automation, dependency and recurring task all worked on the free plan
Pasting a multi-line list creates one task per line
500-row CSV import in under two minutes, and the 500-task board opened in under a second (0.7 s)
Brain² summarised our project accurately, answered "what is overdue?" correctly and asked before adding subtasks
Full CSV export in under a minute: all 771 tasks, with comments and subtasks
Removing someone's access took effect immediately

What falls short

Heavy: about 14 seconds from launch to a usable list and 1.7–2.2 GB of memory
The CSV import matched none of our seven columns automatically and needed status values mapped by hand
The import offers to invite everyone whose email is in the file
Starts with Windows by default and leaves about 275 MB and a broken startup entry after uninstalling
Free AI chat is a one-off trial that does not reset
The export drops dependencies and recurrence settings
Busy setup: nine onboarding screens, a phone-number request, an invite nudge and an AI pop-up for every new list

Scores

Category	Score	Why
Ease of use	6.5	Bulk paste and typed dates are quick; busy setup, an easy-to-misuse priority menu and manual CSV mapping
Views & flexibility	9.0	List, board, calendar and Gantt all free, more a click away
Speed	6.5	Fast imports and a 0.7 s board; ~14 s to become usable, up to 2 GB
Automation & AI	7.5	Rules worked and the AI was accurate and asked first; 100 automation actions/month, non-resetting AI trial
Integrations	7.0	Importers for a dozen tools and a complete CSV export; app integrations not tested
Value for money	8.0	The free plan covered the whole test; paid from $9/user/month billed yearly

Pricing (checked Oct 29, 2026)

Free Forever $0: unlimited tasks and members, 60 MB storage, 5 Spaces, 100 automation actions a month, an AI trial with 500 credits
Core $9/user/month billed yearly
Business $17/user/month billed yearly; premium AI models from Anthropic, OpenAI and Google start here
Enterprise custom

Verdict

ClickUp did everything our test asked on the free plan: four free views, rules that work, a large import that kept every character, a fast 500-task board, and an AI that was right and asked before acting. The costs are weight and friction: a slow start, up to 2 GB of memory, a startup entry added without asking and left behind after uninstalling, and a CSV import that makes you do the matching.

If your team wants one free tool with lists, boards, Gantt charts and automation, ClickUp is a strong choice. If you only need a simple task list, it is more app than you need. See how it compares with Notion, Asana, Linear and Trello in our project management ranking.

BenchTested tests software on a clean Windows PC with the same script for every app. We may earn a commission through some links on the site; it never changes a score. Tested on Oct 29, 2026; each step was run once. We did not test integrations, docs, whiteboards, chat, time tracking, the mobile and Mac apps or paid plans.

6 AI Browsers Tested on Windows 11: Same 8 Tasks, Ranked (2026)

BenchTested — Thu, 29 Oct 2026 20:02:00 +0000

This is a republished copy of our ranking. The original, with a slider to re-weight the scores, a team-cost calculator and all screenshots, lives at benchtested.app/best/ai-browsers.

AI browsers promise an assistant or agent that reads pages, fills forms and carries out multi-step tasks. We installed six of them on a clean Windows 11 virtual machine and gave each the same eight tasks, pasted word for word, on a free test account. The order below comes from those results only.

How we tested

Install from the official site and reach the first agent task; count steps, prompts and minutes, and note anything installed outside the browser.
Run the same 8 scripted tasks: research from official sources, a form fill, a price comparison, a multi-tab extraction, a scheduled routine, a request to buy a book on a demo shop, a five-slide deck and a tricky multi-step form. Each is scored done, partly done or failed, with time and credits used.
Check what the agent may do without asking: purchases, messages, passwords, and whether actions can be reviewed afterwards.
Measure cold start, idle and under-load RAM/CPU against Microsoft Edge on the same 10 tabs.
Check privacy and leftovers: where chats and page data go, and what stays after uninstalling.

Weights: task success ×3, privacy & control ×3, speed & resources ×2, ease of use ×2, value for money ×2, everyday browsing ×1.

The ranking

#	Browser	Score	Best for
1	Polar	7.0	An AI browser that fills forms, opens tabs, builds files and runs scheduled tasks on its own
2	BrowserOS neo	6.3	People who already use Claude Code, Codex or Cursor and want their agent in a separate browser
3	Dia	6.3	Accurate, sourced answers across tabs and an assistant that asks before it acts
4	Strawberry	6.2	Hands-on users comfortable reviewing what it installs
5	Comet	5.2	A normal Chromium browser with a fast answer panel
6	Genspark	5.0	A free agent on a spare profile with no important logins

1. Polar — 7.0/10

Polar's own agent did seven of our eight tasks and half of the eighth, quickly and with honest reports, on a free plan that barely noticed the cost.

Most tasks took 15 to 45 seconds; the five-slide deck about 50 seconds
Our eight tasks used 52 of the 1,000 starting credits, and 50 more are added every day
2.4 GB of memory with our ten test tabs, the same as Edge
No admin rights, no services, no startup entries
But: an account is required even for plain browsing, the agent's prompts and screenshots go to Polar's servers and unnamed AI providers, and it ran on Chromium 153 while Chrome was at 156

Full Polar test log →

2. BrowserOS neo — 6.3/10

No AI of its own: it is a free, open-source (AGPL-3.0) browser built for an outside agent to drive over MCP. With Claude connected it completed all eight tasks and reported every result accurately, but that score belongs to the pair, and you need a paid AI agent and some setup.

Each task runs in its own named tab group; every session is recorded locally and can be replayed
One-click connection for Claude Code; works with Codex, Cursor and other MCP clients
But: listens on two ports on every network interface, telemetry is on by default, and 752 MB of profile data and recordings stay after the default uninstall

3. Dia — 6.3/10

The most careful AI browser we tested: it asks before it reads or opens anything, got every research task right and built a real, sourced slide deck. But it cannot click or type, so both form tasks failed, and it cannot create scheduled routines.

Asks permission with Allow once, Always for this chat or Deny
Uninstalls cleanly in under two seconds
But: onboarding pre-ticks content-data sharing and more, and it adds a firewall rule that allows outbound traffic for any program

4. Strawberry — 6.2/10

Seven of eight tasks fully right, from quick research to a sourced slide deck and an 11-field tricky form.

But: it creates two enabled local Windows accounts via an admin "sandbox helper" that stay after uninstalling, used 5.3 GB of memory for ten tabs against 2.4 GB in Edge, and clicked "Add to basket" without asking

5. Comet — 5.2/10

On the free plan, a fast, well-behaved Chromium browser with a quick answer panel, but not an agent: in eight tasks its Assistant never once clicked, typed or opened a tab.

Answers in 2–5 seconds with sources; 3.0 GB for ten tabs
But: slide decks need Pro, it installs three Windows services for the whole PC, and two updater services stay after uninstalling

6. Genspark — 5.0/10

Its free side-panel agent really does drive the browser: it researched, filled a form and compared prices well.

100 free credits a day covered all eight tasks with 29 to spare
But: built on Chrome 141 while Chrome was at 156, it left two fields empty on the tricky form while reporting them filled, and it clicked "Add to basket" without asking

Verdict

Polar scored highest in our tests. If it doesn't fit: BrowserOS neo for Claude Code/Codex/Cursor users, Dia if you want an assistant that never acts without asking, Strawberry for hands-on users, Comet as a plain browser with fast answers, Genspark on a spare profile.

Head-to-head pages: Polar vs Dia · Polar vs BrowserOS neo

BenchTested tests software on a clean Windows PC with the same script for every app. We may earn a commission through some links on the site; it never changes a score. Each task was run once, on free plans, in October 2026.