Skip to content
Grill · 20 Nov · Round 3 of 5

UncensoredLabs landing page review

69
Cooked

Higher than 38 of the 63 other launches on the public record, level with 3 (60%)

“The page sells a technical capability to everyone, leaving the ideal buyer to find their own reason to pay.” 

What we read, 20 Nov at 19:05: the page, the page its main button opens, at laptop and phone width.

The numbers

This page, in numbers.

Its four scores against the public record, and its public grills over time.

Its four axes

uncensoredlabs.devthe median launch on the public record
81
median 81
64
median 64
100
median 77
25
median 39
Audience252 of 4 tests pass

Security research appears halfway down, but the top speaks to anyone with a product, and the refusal problem comes late.

  • missed:A1Named
  • passed:A2Their words
  • missed:A3Pain before product
  • passed:A4The payer

A1, missed: your product.

Choose an axis to read it.

Where 69 sits

38 of 63other launches score lower

0s10s20s30s40s50s60s70s80s90smedian 66

The public record, 22 Nov: each launch counted once, by its latest public grill.

Its public grills

3 publicsince 26 Oct. This one, 69, is the best.

05010052646926 Oct2 Nov20 Novv0.14

Private grills are left out: only the owner sees them.

First screen

What a stranger sees first.

fine

Before scrolling, on a laptop and on a phone.

predicted, not measured
The page's first screen on a phone, predicted first looks, brighter means more
Phone 390 × 844 · header 64 px
PartScreenTodayFixed
  • Headline9.5%28.3%25.6%
  • Subheadline4.9%10.3%9.3%
  • Menu bar8.2%14.5%19.5%

After five seconds a stranger would say: “Some uncensored AI thing you can chat with or build on, I think.” They should be able to say: “An OpenAI-compatible API for uncensored models, paid per token, with free chat to try.”

Predicted, not measured: a model trained on 1,721 screens people looked at for 3 seconds with eye trackers. On 27 pages it had not seen it ranked the spots people looked at above other spots 83% of the time; “people look at the middle” scores 59%. It can over-weight loud things. The fixes are drawn by us on a copy of your page in our browser; your page is not touched. UEyes (Jiang et al., CHI 2023, CC BY 4.0) · TranSalNet · 20 Nov

Your message

Your message, as a buyer reads it.

4 fixes · 2 to keep

Four questions, in the order a buyer asks them. Each fix quotes your page and ends in something you can do today.

Fatal · they leave before they understandCostly · they understand and still leaveDrag · it works, and could work better

1. Do I get what it is?

Clarity 814 of 5 tests pass.
  1. Fix first · 2CostlyClarity · C2 partialminutesup to +6.7
    “Uncensored AI. Chat. Create. Build.”

    The headline describes the tool's nature, not the result the buyer achieves by using it.

    Proposalyou decide
    Replace the headline with the outcome, e.g., ‘Build AI tools that never refuse a prompt’.

2. Is it for me?

Audience 252 of 4 tests pass.
  1. Fix first · 1FatalAudience · A1 failminutesup to +5.6
    “your product”

    A visitor does not know if this is for security researchers, hobbyists, or enterprises and leaves.

    Proposalyou decide
    Replace ‘your product’ with a named audience, e.g., ‘For security researchers and AI developers’.
  2. Fix first · 3CostlyAudience · A3 failan afternoonup to +5.6
    “Uncensored AI. Chat. Create. Build.”

    The product is presented before the pain of model refusals, missing the emotional hook for the buyer.

    Proposalyou decide
    Move the ‘Refusal-style example’ or a statement about model refusals into the first viewport.

What people type

Yours“uncensored ai”, the phrase in its headline, is the most searched of the five.

“uncensored ai” on Google, week by week

Shaded, the last 3 months: −6% on Google · Bing agrees: −13%

Nov 2025Nov 2026
  • “uncensored ai”64✓ headline
  • “uncensored llm api”under 1no
  • “abliterated model api”under 1no
  • “unfiltered ai api”under 1no
  • “llm api without refusals”under 1no

Google Trends: the five phrases in one search, worldwide, the past 12 months, so they share one scale (100 = the busiest week of any of them; shares, not counts). Bing Webmaster Tools: the exact phrase, worldwide, an average week. Read 20 Nov.

Against this page

Its headline

Uncensored AI. Chat. Create. Build.

The headline uses “uncensored ai”.

3. Do I want it?

Offer 644 of 5 tests pass.
  1. DragOffer · O4 failminutesup to +10.4
    “30 days of usage discounts”

    A discount is a feature; there is no honest reason to buy today rather than bookmarking.

    Proposalyou decide
    Add a launch-specific incentive, such as ‘[number] free credits for the first [number] users’.

4. Do I believe it?

Proof 100Nothing to fix here.
  • Keep: The 'Refusal-style example' provides immediate, concrete proof of the value proposition.
  • Keep: Pricing is transparent and immediately visible in the first viewport.
Why you

Why you, not them.

not answered

Products a buyer may meet while looking for what you sell, read the same day, and the free option most of them pick.

Starts atProof from outsideSays what it is
Abliteration.ai$20 / month (Developer)Logos only: “Made by the team from” Microsoft, BlackRock, Apple, J.P.Morgan; “Backed by” a16z speedrunpartly
Venice$18/mo (Pro); API plans from $68/month (Pro Plus), or pay as you go“Join millions using Venice for private, uncensored AI”; API page: “4M+ users”no

Found by searching your category, read on their own pages the day of this grill. Others a buyer may meet: Featherless, Unfil.AI, unrestricted.sh, NLP Cloud.

The option most buyers pick: Running an abliterated open-weight model on your own machine with Ollama or llama.cpp, or rewording prompts until a mainstream model answers.

A developer who can run a model locally pays nothing per token and keeps every prompt on their own hardware. That is the same privacy promise this page makes.

  1. CostlyWhy younot in the score

    Say why prepaid here beats a monthly plan

    “One-time top-ups from $3. No package required.” and “Credits do not expire.” sit low on the page, under “Choose your package.”

    Similar services start with a monthly plan, and the page never says that it does not need one, which is its clearest difference.

    Proposalyou decide
    No subscription. Top up from $3, and your credits do not expire.
Three visitors

Where each one hesitates.

simulated · not real people

Our critic reads your page as three different strangers would. A simulation from the page alone: not real people, not your analytics.

The AI developer

Builds a security or red-team tool and needs a model that answers every prompt.

Hesitates at the subheadlineA1

“Pay for the tokens your product uses.”

Is this made for security work like mine, or for any app at all?

Would stay if: A first line that names security researchers and AI developers as the people it is for (A1).

The skeptic

Has already paid for another uncensored API and knows the prices are all similar.

Hesitates at the plansO4

“30 days of usage discounts”

2.5% off for a month is not a reason to prepay $20 today.

Would stay if: A clear reason to pick a package now, or a plain line that top-ups from $3 need no package at all (O4).

The newcomer

Has heard of uncensored AI but has never used a model API.

Hesitates at the first screenA3

“One API for uncensored and abliterated models.”

I don’t know what abliterated means, or what problem this fixes for me.

Would stay if: The refusal example ("Request declined") shown near the top, so the problem is visible before the product (A3), under a headline that says what you get (C2).

The pattern: All three have to work out for themselves who the page is for and why it matters now: the audience is unnamed (A1), the refusal problem comes halfway down (A3), the headline lists verbs instead of an outcome (C2), and the packages give no reason to buy today (O4).

Measured, not scored

What passed, and 6 to tidy.

We click your main button, “Try Chat free”, as a visitor would, then read the page as AI and search do, and measure speed, links, accessibility, privacy and the technical basics. Beside the score, never in it.

  1. DragAccessibilitynot in the score
    Fix 3 more accessibility problemsShow the fix
    Elements must only use permitted ARIA attributes · Aside should not be contained in another landmark · Landmarks should have a unique role or role/label/title (i.e. accessible name) combination
    Proposalyou decide
    Fix them in the order listed: each names the element it found.
  2. CostlyPhone speednot in the score

    Unfreeze the page on phones

    Frozen for 1.9 s after it appears; aim for 0.2 s.

    Taps in that time do nothing.

    Proposalyou decide
    Load less JavaScript before the first screen is usable.
  3. DragPrivacynot in the score
    Set cookies only after consentShow the fix
    __cf_ob, cf_clearance

    Worth checking, not a verdict: in the EU most cookies need consent first.

    Proposalyou decide
    Set them only after the visitor agrees.
  4. DragTechnicalnot in the score
    Fix 1 technical settingShow the fix
    DMARC: p=none
    Proposalyou decide
    Each is a setting on your host or DNS, named in the line above.
✓ Everything else passed“Try Chat free” goes to /signup, no account asked first · laptop 1.1 s · 318 KB · no layout shift · 10 of 11 technical checksShow all results
Largest paint, desktop
1.1 s
target ≤ 2.5 s
Layout shift
0
target ≤ 0.1
Links on the page
1 broken
3 checked; 29 could not be checked
Accessibility
3 issues
elements must only use permitted aria attributes; aside should not be contained in another landmark; landmarks should have a unique role or role/label/title (i.e. accessible name) combination
Frozen on a phone
1.9 s
after it appears; taps do nothing. Target ≤ 0.2 s
Weight on a phone
318 KB
69 requests

The main button: “Try Chat free”

Leads to/signup?next=chat
Fieldsnone
A passwordno
Email to confirmno
A cardno

The look

Font families2: Schibsted Grotesk, Cascadia Code
Text sizes27, 8 to 192 px
Filled button styles6
Images far from their shown size0

Search and sharing

Privacy and the law · worth checking, not a verdict

Cookies before any click__cf_ob, cf_clearance
Third parties on loadnone
Imprintno link found
Privacy/privacy · 403
Terms/terms · 403

Technical

HTTPS and redirect301 → https://uncensoredlabs.dev/
certificatevalid to 2027-01-24 (65 days), Google Trust Services
HSTSmax-age=31536000; includeSubDomains
CSPset
frameDENY
nosniffnosniff
referrerstrict-origin-when-cross-origin
permissionsset
security.txt200
SPFv=spf1 include:_spf.mx.cloudflare.net ~all
DMARCp=none
Launch day

Ready for Show HN and Product Hunt?

2 to fix first

What those crowds check first, and drafts written from your page's own words. Edit them before you post; we never post for you.

Can I try it without an account?No, “Try Chat free” leads to a sign-up page first.fix first
Who made it?No person is named on the page.fix first
Does the page say what it is?Yes, the first screen says “One API for uncensored and abliterated models.”ready
Is the price shown?Yes, per-token rates like “$1.00 / 1M in” and packages from $20 are on the page.ready
Drafts for Show HN, Product Hunt and X“Show HN: UncensoredLabs – OpenAI-compatible API for abliterated models”Show the drafts
Show HN titleShow HN: UncensoredLabs – OpenAI-compatible API for abliterated models
Product Hunt taglineOne API for uncensored and abliterated models
First commentI built UncensoredLabs because [your reason]. It is one OpenAI-compatible API for uncensored and abliterated models, such as GLM-5.3 Abliterated and DeepSeek V4.1 Flash 1017 Abliterated. You keep your OpenAI client, change the base URL, and pay per token from prepaid credits that do not expire. [your question]
X postUncensoredLabs: one API for uncensored and abliterated models. Keep your OpenAI client, change the base URL, pay per token. GLM-5.3 Abliterated from $1.00 per 1M input tokens. Top-ups from $3, and credits do not expire. https://uncensoredlabs.dev/
The score

Your score in detail.

How 69 adds upevery axis and its weight, the level, every test, every grill of this pageShow the sum
Clarity35%
81
+28.4
Audience18%
25
+4.5
Offer29%
64
+18.6
Proof18%
100
+18.0
The sum 69.4, so
69
11 points to Well done

Level 3 of 4 Cooked

all open: 100
now 69
0406080
4
left to fix
+31
points in reach
100
could reach · Well done
Badges
  • ✓First grill
  • ✓Cooked — reached 60
  • Well done — no round left on this launch
  • ✓Before and after — finish round 2
Every point from a test

The four axes

Every point comes from a test. Open a row, or read the rubric.

Clarity35% of the score
81
C1
One API for uncensored and abliterated models.
pass
C2
Uncensored AI. Chat. Create. Build.
partial
C3
Headline is free of adjectives like 'seamless' or 'powerful'.
pass
C4
Focuses solely on uncensored AI API access.
pass
C5
Try Chat free
pass
4 passed · 1 partial · 0 failed → 81. Weighted at 35%, this contributes 28.4 of the overall 69.
Audience18% of the score
25
A1
your product
fail
A2
abliterated
pass
A3
The product is stated in the first line; the pain of refusals is shown halfway down the page.
fail
A4
Pay for the tokens your product uses.
pass
2 passed · 0 partial · 2 failed → 25. Weighted at 18%, this contributes 4.5 of the overall 69.
Offer29% of the score
64
O1
$1.00 / 1M in
pass
O2
Clear distinction between one-time top-ups and Developer/Growth/Scale packages.
pass
O3
Token pricing is consistent with industry standards for similar models.
pass
O4
30 days of usage discounts
fail
O5
Try Chat free
pass
4 passed · 0 partial · 1 failed → 64. Weighted at 29%, this contributes 18.6 of the overall 69.
Proof18% of the score
100
P1
90.3B Tokens processed
pass
P2
935.5K Successful requests
pass
P3
Dashboard screenshot and cURL example.
pass
P4
Specific usage stats match the scale of the offering.
pass
4 passed · 0 partial · 0 failed → 100. Weighted at 18%, this contributes 18.0 of the overall 69.
28.4 + 4.5 + 18.6 + 18.0 = 69.4 → 69.
Full page read.
3 grills

Every grill of this page

  1. 52Undercooked26 Oct 2026
  2. 64Cooked2 Nov 2026+12
  3. 69Cooked20 Nov 2026+5You are here

Your launch, read the same way.

Free. No card.

Ready to launch? The Launch Kit: your agent submits you to up to 85 free directories · $49 →