Banja
About
Services
Products
Case Studies
Lab
Contact Us
Let us pitch to you

LET'S BUILD
THE FUTURE.

Start a Project
or
Meet Jett
banja.au

We build digital products for people who move fast.

Explore

•About•Case Studies•Blog•Careers•Contact

Services

•Product Design & Build•AI Agents & Automation•Website & Brand Setup

Products

•Boosta

Contact

helloremovethis@andthisbanja.au
50 Miller St
North Sydney NSW 2060

© 2026 Banja Labs. All rights reserved.

Privacy PolicyTerms of Use

Banja Lab / Benchmarks / Test

WDB-0004webdesign · medium

The Gross Lab children's museum exhibition companion site

The same task, run on 28 models. Compare the outputs side by side, or open any one in a popup to inspect it.

Top result: Ingrid Kessler (claude-fable-5) (high reasoning) at 96.9% composite. Lowest: Sol Kepler (deepseek-v4-pro) at 37.5%. 28 models compared on this task.

How it ran
  • Each model was given the brief below in a fresh, isolated session with no access to our tools, and returned a single self-contained index.html (inline CSS and JS, no external requests, no build step).
  • The rendered output was scored 1 to 5 on brief fidelity, visual design, craft, and impact by a four-family vision panel - Anthropic (Claude Opus 4.8), OpenAI (GPT-5.5), Google (Gemini 3.1 Pro), and xAI (Grok 4.3) - using one identical prompt so the scores compare. The published judge score is leave-one-family-out: a model is never scored by a judge of its own family, so same-family self-preference is removed.
The brief

# Client brief: The Gross Lab You are designing and building the companion website for a real touring museum exhibition. The brief below gives you the business facts, the goals, the audiences, the sections the site must have, and the exact content to use. It gives you no visual direction on purpose. Every design decision - layout, type, colour, imagery, motion, tone - is yours to make. Read the whole brief, form a point of view on how this exhibition should look and feel, and build it. ## The commissioning body The Riverbend Discovery Centre is a regional children's science museum in Albury, New South Wales. It runs hands-on galleries for primary-school-aged visitors and takes bookings from schools across the Riverina and north-east Victoria. Its touring exhibition, The Gross Lab, is about the biology of disgusting things: what your body makes, why it makes it, and what eats what when something dies. The exhibition is built for children aged 7 to 12 and opens at the Centre on 19 September 2026, running until 25 January 2027 before it tours to other regional centres. This site is the exhibition's companion: it is where a child gets excited before or after a visit, and where a parent or a teacher decides the exhibition is worth the trip and books it. ## What the site is for 1. Make a 7 to 12 year old want to come, by telling them the actual science of the six exhibits rather than teasing them with it. 2. Convince a parent or a teacher that this is a serious, safe, curriculum-linked science exhibition and not a shock show. 3. Bring in bookings: sell general tickets for the four daily sessions, and get teachers to submit the school group booking form. ## Who it is for Two audiences on one page, and both must be served properly: - Children aged 7 to 12, many of whom will read the site themselves on a phone or a school laptop. - The adults who bring them: parents, grandparents, carers, and primary-school teachers booking a class of twenty-five. ## The experience the Centre wants The Centre has one request about the experience, not about the look. Their exhibition rule, printed on the wall of the gallery, is: **make them laugh, then make them learn.** The laugh is the hook; the science is the point. Nothing in the exhibition is disgusting for its own sake, and the Centre is firm that the site must not be either. Every joke should be attached to something true, and no child should leave the page with a wrong idea about how their body works. This is a note about content and tone, not a style instruction. The site must also carry an interactive element - the gross-o-meter - described in the content below. How it behaves is your call. One firm rule: any motion must respect a visitor's reduced-motion preference. When reduced motion is requested, the whole site must still be completely readable, every section reachable, and the gross-o-meter still usable as a calm, static control, with no information conveyed only through movement. ## What the site must contain (all of these sections) - a hero that opens the exhibition, - what the exhibition is, - the six exhibits, - the gross-o-meter, an interactive element where a visitor rates or reveals how gross each exhibit is, - the real science explainer, - a parents and teachers section covering safety, curriculum links, and accessibility, - session times and ticket prices, - school bookings, with the booking form, - visit and contact details. ## Content to use Use the content below substantially as written, in Australian English. You may write your own section headings and any short connective microcopy (button labels, captions, form labels, gross-o-meter prompts), but the facts, figures, names, numbers, prices, and descriptions are fixed. Do not invent extra exhibits, drop any of the six, or change the science, the ratings, the prices, the session times, the curriculum links, the accessibility notes, or the contact details. ### The exhibition The Gross Lab is a touring exhibition about the biology of disgusting things, made for children aged 7 to 12. It has six exhibits, takes about 60 minutes to walk through, and is staffed by explainers in lab coats who will answer anything. Every exhibit pairs one genuinely revolting thing your body or the natural world does with the reason it does it. The exhibition opens at the Riverbend Discovery Centre on 19 September 2026 and runs until 25 January 2027. An exhibition ticket includes entry to the rest of the Centre for the day. ### The six exhibits Present all six. Each has a kid-facing explanation, one surprising fact with a number, a one-line note for adults, and the Lab's own gross rating out of 5. **1. Snot Factory** - Lab rating 3 out of 5. Kid-facing: Your nose and airways make mucus all day long, and it is sticky on purpose. Dust, pollen and germs get stuck in it before they can reach your lungs. Tiny hairs called cilia then sweep the whole sticky load up to your throat, and you swallow it without ever noticing. Surprising fact: your body makes about 1.5 litres of mucus a day, roughly a large milk bottle, and you swallow nearly all of it. For adults: mucus is a first line of immune defence, so children learn to read a runny nose as the body working rather than failing. **2. Scab Squad** - Lab rating 2 out of 5. Kid-facing: When you cut your knee, cells called platelets rush to the hole and stick together into a plug. A stringy protein called fibrin then weaves a net across the plug, the net dries out, and that hard brown lid is your scab. Underneath it, brand new skin is being built. Surprising fact: platelets reach a fresh cut in under 30 seconds, and a normal clot is finished in about 5 minutes. For adults: an introduction to blood components and clotting, and a practical reason not to pick a scab. **3. The Acid Pit** - Lab rating 2 out of 5. Kid-facing: Your stomach makes hydrochloric acid, the same acid used in school science labs. It turns your lunch into mush and kills most of the germs that came along with it. Your stomach does not dissolve itself because it is constantly rebuilding its own slimy lining. Surprising fact: stomach acid runs at about pH 1.5, and your stomach grows itself a whole new lining roughly every 4 days. For adults: pH, chemical digestion, and a safe first step into chemistry. **4. Dung Beetle Depot** - Lab rating 4 out of 5. Kid-facing: Dung beetles find fresh animal poo, roll it into a ball, and bury it underground for their babies to eat. Doing that clears the paddock, feeds the soil, and stops flies breeding in the pat. Australia had to bring dung beetles in from overseas, because our native ones handle kangaroo droppings and will not touch a cattle pat. Surprising fact: one cow drops about 12 pats a day, and a working crew of dung beetles can bury a whole pat in under 48 hours. For adults: ecosystem services, decomposition, and a clear Australian example of biological control. The Depot holds a live, sealed dung beetle colony. **5. The Forensic Log** - Lab rating 5 out of 5. Kid-facing: When an animal dies, blowflies find it within minutes and lay their eggs. The eggs hatch into maggots that eat the soft parts and grow astonishingly fast, and a scientist can measure a maggot to work out how long the animal has been dead. Nothing in nature is wasted, because decomposers return every dead thing to the soil. Surprising fact: one blowfly can lay about 250 eggs at a time, and its maggots reach full size in roughly 5 days. For adults: decomposition, food webs and nutrient cycling, taught with models and a time-lapse of a fallen log. There are no human remains and no photographs of real bodies. **6. Pong Patrol** - Lab rating 3 out of 5. Kid-facing: Sweat is almost entirely water and salt, and on its own it does not smell at all. The smell comes from bacteria that live on your skin, especially in warm spots like armpits, feeding on your sweat and giving off waste gases. So body odour is really just bacteria burps. Surprising fact: you have about 3 million sweat glands, and on a hot day you can lose more than a litre of sweat in an hour. For adults: thermoregulation and the skin microbiome, and a gentle way into a hygiene conversation before high school. ### The gross-o-meter Every exhibit in the gallery has a gross-o-meter dial on its panel, and visitors vote on the way out. The scale has five named levels, and the site must use these names: 1. Barely Icky 2. Bit Yuck 3. Proper Gross 4. Full Squirm 5. Absolutely Revolting The site must carry the gross-o-meter as a working interactive element: a visitor rates or reveals how gross each of the six exhibits is, and can compare their rating against the Lab's own rating for that exhibit (Snot Factory 3, Scab Squad 2, The Acid Pit 2, Dung Beetle Depot 4, The Forensic Log 5, Pong Patrol 3). Design the behaviour yourself. It must work with a keyboard and on a touchscreen, and nothing is submitted anywhere: the ratings stay in the page. ### The real science explainer Use this content for the explainer: Disgust is not a joke to scientists. It is a studied human emotion that evolved to keep us away from things that make us sick, which is why a rotten smell turns your stomach before you have decided anything about it. The Gross Lab uses that reaction as a hook, then explains the biology behind it. Every panel in the exhibition was written by the Centre's education team and reviewed for accuracy by Dr Priya Raman, a microbiologist. The exhibition was trialled with 240 students from six schools before the tour began. A fortnight after their visit, 9 in 10 of those students could still explain at least one of the six science concepts in their own words. Nothing in The Gross Lab is disgusting for its own sake. If a thing is in the exhibition, there is real biology underneath it. ### Parents and teachers Safety and content notes: - Nothing on display is human tissue. The specimens are models, casts, and preserved invertebrates. - The dung beetle colony is live and sealed, and cannot be opened by visitors. - There are no jump scares, no loud surprises, and no gore. - The Forensic Log has a signposted bypass door for children who would rather skip the maggots, and it rejoins the gallery at the next exhibit. - There is a hand-washing station at the gallery exit, and explainers are stationed at every hands-on table. Curriculum links. The exhibition is mapped to two strands of the Australian Curriculum (Version 9) Science learning area: - Science Understanding, Biological sciences (AC9S4U01) - explain the roles and interactions of consumers, producers and decomposers within a habitat and how food chains represent feeding relationships. - Science as a Human Endeavour, Use and influence of science (AC9S6H02) - investigate how scientific knowledge is used by individuals and communities to identify problems, consider responses and make decisions. A four-page teacher pack with pre-visit and post-visit activities for each of the six exhibits is emailed with every confirmed school booking. Accessibility: - Step-free and wheelchair accessible throughout, and every hands-on table is at seated height. - A sensory-friendly session runs on the first Sunday of each month at 9:00 am, with lower lighting, sound off, and capped numbers. - A quiet room with low lighting and beanbags is open all day beside the gallery entrance, and visitors can come and go from it freely. - An Auslan-interpreted session runs on the third Saturday of each month at 11:15 am. - Companion Cards are accepted, and the companion enters free. ### Sessions and tickets The Gross Lab runs four sessions a day, each about 60 minutes: - 9:45 am - 11:15 am - 1:00 pm - 2:30 pm Ticket prices (each ticket includes entry to the rest of the Centre for the day): - Child, ages 5 to 16 - $14 - Adult - $18 - Family, 2 adults and up to 3 children - $52 School groups: $9 per student, with one teacher free for every 10 students. School sessions run Tuesday to Friday during term at 9:45 am and 11:15 am, and must be booked at least 2 weeks ahead. Children under 5 enter free but the exhibition is written for ages 7 to 12. ### School bookings Teachers book through the form on this page. The form has exactly these fields: - School name - Contact name - Email - Year level - Number of students - Preferred date The Centre's bookings officer, Marcus Ellery, replies within 2 business days to confirm a session and send the teacher pack. The form does not need a working backend: on submit, acknowledge the teacher in the page. ### Visit and contact Riverbend Discovery Centre 18 Willow Bend Road, Albury NSW 2640 Phone (02) 6021 4488 Email hello@riverbenddiscovery.org.au School bookings: schools@riverbenddiscovery.org.au Open Tuesday to Sunday, 9:30 am to 4:30 pm. Closed Mondays outside school holidays, and closed Christmas Day and Good Friday. Free parking on site, and it is a 10 minute walk from Albury station. ## Output contract (React, single page) Build one single-page site. Output exactly two files: - `App.tsx` - a default-exported React component. Define any subcomponents in this same file. This is the whole page. - `theme.css` - a Tailwind v4 `@theme` token block, your `font-family` choices from the shared font menu below, and any custom CSS you need. The only imports allowed in `App.tsx` are `react`, `framer-motion`, `lucide-react` (never the Sparkles icon), and, for 3D, `three`, `@react-three/fiber`, and `@react-three/drei`. No other packages, no network requests, and no external assets: all imagery must be inline SVG, CSS, drawn in code on a canvas, or rendered with WebGL. The fonts are bundled offline, so use only the shared menu: Archivo (including Archivo Expanded), Instrument Sans, Space Grotesk, Work Sans, IBM Plex Sans, IBM Plex Mono, JetBrains Mono, Fraunces, Instrument Serif, Playfair Display, Lora, Caveat. Make it responsive across mobile, tablet, and desktop. The page must also remain fully readable and usable if WebGL is unavailable: never leave a visitor with a blank screen or a missing section because a 3D canvas failed to start.

Ingrid Kessler (claude-fable-5)
High reasoning
Ingrid Kessler (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 96.9%, judge 4.9/5
Composite 96.9%Judge 4.9/5
Judge panelAnthropic 4.8/5OpenAI 5.0/5
single-judge (Claude) 4.8/5 → leave-one-family-out 4.9/5

Anthropic: Every required section is present and correct: hero ("Made for ages 7-12"), what-it-is, all six exhibits with kid facts, numbers and Lab ratings out of 5, the gross-o-meter as a "You be the judge" interactive with 1-5 buttons and a reveal-the-Lab's-rating control, the science explainer with the 240-students and 9-in-10 stats and Dr Priya Raman, parents/teachers covering safety, accessibility and curriculum links, four sessions with correct $14/$18/$52 prices, the six-field school booking form, and full contact details. The visual language is bold, cohesive and on-theme - a dark playful hero, l

OpenAI: The page visibly includes all required sections, the six exhibits with correct science, ratings, sessions, prices, school form fields, safety/curriculum/accessibility notes, and contact details. The design is highly cohesive and polished, with bold kid-friendly branding balanced by credible adult-facing information, strong layout rhythm, and distinctive colour/type choices that make the result memorable.

Full run
Mara Vieira (kimi-k3)
default reasoning
Mara Vieira (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 96.9%, judge 4.9/5
Composite 96.9%Judge 4.9/5
Judge panelAnthropic 4.8/5OpenAI 5.0/5
single-judge (Claude) 4.8/5 → leave-one-family-out 4.9/5

Anthropic: All required sections are present and correct: hero, what-it-is, the six exhibits (numbered 01-06 with kid facts, number-stats, adult notes and Lab ratings), a working-looking gross-o-meter table, the science explainer with the 9-in-10 stat, safety/curriculum/accessibility, the four session times and $14/$18/$52/$9 prices, a six-field school booking form, and visit/contact details. The design is a bold, coherent brutalist exhibition-poster system - acid lime, red and black with oversized Archivo Expanded display type contrasted against serif italics - that is distinctive and on-brief for a gro

OpenAI: The screenshot shows all required sections with the fixed facts, exhibit content, ratings, safety/curriculum/accessibility notes, tickets, booking form fields, and contact details represented accurately. The bold editorial system, acidic palette, strong typography, and consistent card language make it feel distinctive and highly polished while staying playful rather than gratuitously gross.

Full run
Elias Byrne (Opus 4.8)
High reasoning
Elias Byrne (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 93.8%, judge 4.8/5
Composite 93.8%Judge 4.8/5
Judge panelAnthropic 4.8/5OpenAI 4.8/5
single-judge (Claude) 4.8/5 → leave-one-family-out 4.8/5

Anthropic: Every required section is present and correct: a bold hero, "what it is", all six exhibits with their numbers and Lab ratings, a gross-o-meter with per-exhibit comparison, the science explainer (240 students, 9 in 10), a dedicated parents/teachers "serious page" covering safety, curriculum links and accessibility, four sessions with the exact $14/$18/$52 pricing, school-group pricing, a six-field booking form, and full visit/contact details. The design is a confident editorial/brutalist system - massive Archivo-Expanded display type, an olive-and-lime palette punctuated with dark and yellow-bl

OpenAI: The page includes the required hero, exhibition overview, six exhibits, gross-o-meter, science explainer, parent/teacher information, ticketing, school form, and contact details with the key facts largely correct. The main fidelity weakness is that the adult notes for each exhibit and the gross-o-meter comparison appear hidden in the default screenshot state, so not all required content is visibly available. Visually it is highly cohesive, distinctive, polished, and memorable, with strong typography, illustration, colour, spacing, and section rhythm.

Full run
Mara Vieira (claude-fable-5)
High reasoning
Mara Vieira (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 93.8%, judge 4.8/5
Composite 93.8%Judge 4.8/5
Judge panelAnthropic 4.5/5OpenAI 5.0/5
single-judge (Claude) 4.5/5 → leave-one-family-out 4.8/5

Anthropic: Every required section is present and correct: hero, exhibition intro, all six exhibits with their custom SVG illustrations and gross ratings, a gross-o-meter compare grid, the science explainer with the 240/9-in-10 stats, a parents-and-teachers block split into safety/curriculum/accessibility, the four session times, the three ticket prices, a full school booking form, and the Albury visit/contact details. The dark near-black-and-lime palette with bespoke illustrations for each exhibit is coherent, modern and genuinely distinctive, hitting the "make them laugh, then make them learn" tone whil

OpenAI: The page visibly includes every required section with the fixed exhibits, ratings, science explanations, adult notes, safety/curriculum/accessibility information, sessions, prices, school form, and contact details. The visual system is highly coherent and distinctive, with strong dark-and-slime colour direction, custom illustrations, polished cards, and clear hierarchy. Overall it feels finished, memorable, and well suited to the “make them laugh, then make them learn” tone.

Full run
Elias Byrne (claude-fable-5)
High reasoning
Elias Byrne (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 93.8%, judge 4.8/5
Composite 93.8%Judge 4.8/5
Judge panelAnthropic 4.5/5OpenAI 5.0/5
single-judge (Claude) 4.5/5 → leave-one-family-out 4.8/5

Anthropic: Every required section is present and correct: bold hero, "what it is", all six exhibits with kid facts, numbers, adult notes and out-of-5 ratings, an interactive gross-o-meter grid, the science explainer with the 240 and 9-in-10 stats, parents/teachers (safety, curriculum links, accessibility), session times, ticket prices ($14/$18/$52), the six-field school booking form, and full visit/contact details. The visual system is confident and cohesive - dark hero with oversized Archivo-Expanded display type, a wavy divider into warm cream body, and disciplined green/pink/lime accents with playful

OpenAI: The page visibly covers every required section with the fixed facts, six exhibits, gross-o-meter, adult reassurance, tickets, school form, and contact details all presented accurately and clearly. The visual system is highly coherent and distinctive, with bold child-friendly typography, strong colour coding, polished cards, and a tone that balances gross fun with credible science. Craft is strong throughout, with consistent spacing, hierarchy, iconography, and responsive-looking section composition, creating a memorable premium result.

Full run
Cass Werner (Opus 4.8)
High reasoning
Cass Werner (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 93.8%, judge 4.8/5
Composite 93.8%Judge 4.8/5
Judge panelAnthropic 4.8/5
single-judge (Claude) 4.8/5 → leave-one-family-out 4.8/5

Anthropic: Every required section is present and correct: hero with the "make them laugh, then make them learn" line, the what-it-is strip, all six exhibits with kid facts and gross ratings, a named gross-o-meter voting block (Barely Icky through Absolutely Revolting), the science explainer with the 240/9-in-10/6 stats, parents-and-teachers split into safety, curriculum and accessibility, four session times, the $14/$18/$52 tickets, the six-field booking form, and full contact details. The visual system is coherent and confident - warm paper background, deep-green header and footer, tasteful accent green

Full run
Yuki Tanabe (Opus 4.8)
High reasoning
Yuki Tanabe (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 90.6%, judge 4.6/5
Composite 90.6%Judge 4.6/5
Judge panelAnthropic 4.3/5OpenAI 5.0/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.6/5

Anthropic: Every required section is present and content-accurate: hero, what-it-is, all six numbered exhibits with ratings and illustrations, a gross-o-meter comparison control, the "disgust is not a joke" explainer with the 240 / 9-in-10 / reviewed stats, a parents-and-teachers block covering safety, curriculum and accessibility, the four session times with correct $14/$18/$52 and $9 school pricing, a full booking form, and Albury contact details. The visual language is confident and on-brief - a bold lime-on-dark scheme with playful yet clinical SVG specimens and strong Archivo-style display type that

OpenAI: The page visibly includes every required section with the fixed facts, ratings, prices, session times, curriculum/accessibility notes, booking form fields, and contact details, and the gross-o-meter is clearly represented as an in-page control. The visual system is bold, cohesive, child-friendly without feeling unserious, with strong typography, custom illustrations, consistent cards, and a distinctive museum-quality energy.

Full run
Mara Vieira (Opus 4.8)
High reasoning
Mara Vieira (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 90.6%, judge 4.6/5
Composite 90.6%Judge 4.6/5
Judge panelAnthropic 4.5/5OpenAI 4.8/5
single-judge (Claude) 4.5/5 → leave-one-family-out 4.6/5

Anthropic: Every required section is present and correct: hero, what-it-is, all six exhibits with ratings/facts/adult notes, a working gross-o-meter table using the five named levels with Lab-rating comparison, the science explainer with its 240/9-in-10 stats, parents-and-teachers split into safety/curriculum/accessibility, four session times and the exact ticket prices, the school booking form with the right fields, and full contact details. The design is coherent and modern - a confident dark-green/lime/cream palette, strong Archivo Expanded display type, a petri-dish hero motif and playful torn-paper

OpenAI: The screenshot shows all required sections, all six exhibits with facts/adult notes/ratings, gross-o-meter controls, ticket/session details, school booking form, and visit/contact information with the fixed content largely intact. The neon-on-black lab aesthetic with cream science-card sections is distinctive, coherent, and well composed for a long informational page. Craft is very high, though some small copy and micro-labels feel dense or low-contrast, keeping it just short of flawless.

Full run
Noor El-Amin (Opus 4.8)
High reasoning
Noor El-Amin (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 90.6%, judge 4.6/5
Composite 90.6%Judge 4.6/5
Judge panelAnthropic 4.5/5OpenAI 4.8/5
single-judge (Claude) 4.5/5 → leave-one-family-out 4.6/5

Anthropic: Every required section is present and correct: hero, what-it-is, all six exhibits with kid text, numbered surprising facts, adult notes and out-of-5 ratings, the gross-o-meter with the five named levels and per-exhibit comparison, the science explainer with the 240-students / 9-in-10 stats, parents-and-teachers safety/curriculum/accessibility, session times and the $14/$18/$52 prices, a full booking form with the exact fields, and the visit/contact footer. The visual system is cohesive and confident - a dark petri-dish hero, cream body, acid-green accent, and bold Archivo-style display type th

OpenAI: The page visibly covers almost all required sections and fixed facts, including the six exhibits, gross-o-meter UI, science explainer, parent/teacher information, sessions, prices, booking form, and contact details; the main visible gap is that the required adult notes for each exhibit appear collapsed rather than readable in the screenshot. The visual system is cohesive, playful, and highly polished, with distinctive custom illustrations, strong hierarchy, confident colour, and a premium museum-exhibition feel.

Full run
Yuki Tanabe (claude-fable-5)
High reasoning
Yuki Tanabe (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 90.6%, judge 4.6/5
Composite 90.6%Judge 4.6/5
Judge panelAnthropic 4.3/5OpenAI 5.0/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.6/5

Anthropic: Every required section is present and correct: hero, what-the-exhibition-is, all six exhibits with kid facts/numbers/adult notes/ratings, a gross-o-meter block with the six named levels, the science explainer with the 240-students and 9-in-10 stats, a parents/teachers block covering safety, curriculum links and accessibility, the four session times, correct ticket prices ($14/$18/$52), the school booking form with the right fields, and the Albury contact details. The visual system is coherent and on-theme - a dark olive base with a lime specimen-jar accent and warm cream exhibit cards reads pl

OpenAI: The page visibly includes all required sections, all six exhibits with their science, ratings, adult notes, the gross-o-meter controls, parent/teacher detail, tickets, school form, and contact information with the specified facts preserved. The visual system is highly coherent and distinctive, with strong typography, playful scientific illustration, excellent spacing, and a polished gross-but-not-gory tone that fits the brief.

Full run
Ingrid Kessler (Opus 4.8)
High reasoning
Ingrid Kessler (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 84.4%, judge 4.4/5
Composite 84.4%Judge 4.4/5
Judge panelAnthropic 4.3/5OpenAI 4.5/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.4/5

Anthropic: Every required section is present and correct: oversized hero, exhibition intro, all six exhibits with kid explanations, numbered surprising facts, adult notes and Lab ratings, a named gross-o-meter comparison section, the science explainer with 240/6/9-in-10 stats, the grown-ups block covering safety/curriculum/accessibility, session times, ticket prices, the six-field school booking form, and full contact details. The dark editorial treatment with a lime accent over playful pastel exhibit cards is coherent, modern and distinctive, with strong type hierarchy and a confident hero. Craft is cle

OpenAI: The page is highly polished, distinctive, and visually coherent, with strong typography, colour, illustration, and section rhythm that fits the exhibition tone well. Most required content is present, including all six exhibits, the gross-o-meter, science explainer, grown-up notes, booking form, and contact details, but the visible page appears to omit the dedicated general session times and ticket prices section. Craft is strong overall, with only minor issues such as tight tiny text and awkward email wrapping in the footer.

Full run
Noor El-Amin (claude-fable-5)
High reasoning
Noor El-Amin (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 84.4%, judge 4.4/5
Composite 84.4%Judge 4.4/5
Judge panelAnthropic 4.0/5OpenAI 4.8/5
single-judge (Claude) 4.0/5 → leave-one-family-out 4.4/5

Anthropic: Every required section is present and correct: hero, what-it-is, all six exhibits with ratings/facts/adult notes, a "You Be The Judge" gross-o-meter block, the science explainer with the 240-students and 9-in-10 stats, parents/teachers with safety/accessibility/curriculum links, session times and the $14/$18/$52 prices, the school booking form with the exact fields, and the Albury visit/contact details. The lime-green, cream and black palette is coherent and on-theme for a playful science lab, with strong bold display type and a consistent card system that reads as finished across the alternat

OpenAI: All required sections and fixed facts are visible, including the six exhibits with ratings, gross-o-meter controls, science explainer, parent/teacher notes, tickets, school booking form, and contact details. The design is highly coherent and polished, with strong typography, careful spacing, playful colour, and consistent cards/icons. It feels premium and memorable, though its impact is more refined editorial than truly show-stopping.

Full run
Noor El-Amin (kimi-k3)
default reasoning
Noor El-Amin (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 84.4%, judge 4.4/5
Composite 84.4%Judge 4.4/5
Judge panelAnthropic 4.0/5OpenAI 4.8/5
single-judge (Claude) 4.0/5 → leave-one-family-out 4.4/5

Anthropic: All required sections are present and content-rich: hero, what-it-is, all six exhibits with ratings/facts/adult notes, an interactive gross-o-meter with per-exhibit rating rows, the science explainer with the 240-student and 9-in-10 stats, parents/teachers (safety, curriculum links AC9S4U01/AC9S6H02, accessibility), the four sessions with $14/$18/$52 pricing and school rates, a six-field booking form, and full visit/contact details. The warm cream palette with serif display headings is coherent, tasteful, and well-spaced, and the cards are cleanly and consistently built. The main weakness is i

OpenAI: The page visibly includes all required sections and fixed facts: six exhibits with ratings and science, gross-o-meter controls, adult reassurance, tickets, school form, and contact details. The visual system is cohesive, polished and well spaced, with charming hand-drawn science motifs and strong typography. Its restraint feels premium and trustworthy, though the gross/kid appeal is a little gentle rather than truly arresting.

Full run
Sol Kepler (Opus 4.8)
High reasoning
Sol Kepler (Opus 4.8) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 81.3%, judge 4.3/5
Composite 81.3%Judge 4.3/5
Judge panelAnthropic 4.3/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.3/5

Anthropic: Exceptionally complete against the brief: all nine sections present and correct, all six exhibits with facts/ratings, the gross-o-meter using the five mandated level names with reveal/reset controls, correct prices, curriculum codes, stats, and the full booking form. The warm-cream / dark-green / lime palette is coherent and strikes the intended kid-but-serious tone, with consistent card and section rhythm throughout. Craft is finished and clean; it stops short of top marks on impact because the type and mid-page layout, while polished, are more competent-editorial than genuinely arresting.

Full run
Cass Werner (claude-fable-5)
High reasoning
Cass Werner (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 81.3%, judge 4.3/5
Composite 81.3%Judge 4.3/5
Judge panelAnthropic 4.3/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.3/5

Anthropic: Every required section is present and correct: hero, what-it-is with the 60-min/7-12/ratings callout, all six exhibits with kid facts, numbered surprising stats, adult notes and out-of-5 badges, a full Gross-o-Meter panel listing all six exhibits against the five named levels, the science explainer with the 240-students/9-in-10 stat tiles, a three-column parents/teachers block (safety, curriculum links, accessibility), correct session times and $14/$18/$52 pricing, the six-field school booking form, and complete contact/hours details. The green-and-cream palette reads on-theme (lab/biology) wi

Full run
Sol Kepler (claude-fable-5)
High reasoning
Sol Kepler (claude-fable-5) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 81.3%, judge 4.3/5
Composite 81.3%Judge 4.3/5
Judge panelAnthropic 4.3/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.3/5

Anthropic: Every required section is present and correct: hero, what-it-is with stat tiles, all six exhibits with ratings/facts/adult notes, an interactive gross-o-meter using the five named levels, the science explainer with Dr Priya Raman and the 240-students/9-in-10 stats, parents-and-teachers split into safety/curriculum/accessibility, four session times with the $14/$18/$52 prices and school-group pricing, a bookings form with the correct fields, and the Albury visit/contact details plus opening hours. The design is coherent and age-appropriate, using bold Archivo-style type and a confident block-co

Full run
Ingrid Kessler (kimi-k3)
default reasoning
Ingrid Kessler (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 78.1%, judge 4.1/5
Composite 78.1%Judge 4.1/5
Judge panelAnthropic 4.3/5OpenAI 4.0/5
single-judge (Claude) 4.3/5 → leave-one-family-out 4.1/5

Anthropic: Every required section is present and correctly populated: hero, what-it-is with stats, all six exhibits with kid facts, adult notes and ratings, a gross-o-meter block using the five named levels with per-exhibit pick prompts, the science explainer with Dr Priya Raman and the 240-student stat, parents/teachers (safety, curriculum codes, accessibility), four session times with the $14/$18/$52 prices, a school booking form with the exact fields, and full visit/contact details. The execution is a confident, cohesive editorial system - large Archivo display type, hairline rules, small-caps section

OpenAI: The page visibly includes all required sections, all six exhibits with correct ratings and science content, the gross-o-meter, adult safety/curriculum/accessibility details, tickets, school form, and contact information. The design is coherent, polished, and editorial, with strong hierarchy and spacing, but it feels very austere and adult-facing for a 7–12 children’s gross-science exhibition. Craft is consistent overall, though the gross-o-meter presentation is a little unclear visually and the overall result lacks the playful, memorable impact the brief could support.

Full run
Elias Byrne (kimi-k3)
default reasoning
Elias Byrne (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 78.1%, judge 4.1/5
Composite 78.1%Judge 4.1/5
Judge panelAnthropic 4.0/5OpenAI 4.3/5
single-judge (Claude) 4.0/5 → leave-one-family-out 4.1/5

Anthropic: Exceptionally thorough brief coverage: every required section is present and correct, including all six exhibits with kid facing text, numbered surprising facts, adult notes and green square gross ratings, the gross-o-meter with the five named levels and per exhibit Lab ratings, the full science stat block, curriculum links, all form fields, and complete contact/hours details. The dark technical editorial layout is coherent, well composed and cleanly crafted with a strong grid, consistent hairlines and tidy tables. The main weakness is tone/impact: the austere monochrome data catalogue look re

OpenAI: The page visibly includes all required sections, the six exhibits with correct ratings and facts, the gross-o-meter controls, adult guidance, tickets, booking form fields, and visit details. The visual system is coherent and polished with a strong dark lab-console aesthetic, though it is quite text-heavy and austere for 7–12 year olds, with limited playful or memorable gross-world imagery. Craft is consistent and finished overall, but the dense small type and restrained presentation reduce excitement and accessibility impact slightly.

Full run
Yuki Tanabe (kimi-k3)
default reasoning
Yuki Tanabe (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 75.0%, judge 4.0/5
Composite 75.0%Judge 4.0/5
Judge panelAnthropic 4.0/5OpenAI 4.0/5
single-judge (Claude) 4.0/5 → leave-one-family-out 4.0/5

Anthropic: Brief fidelity is strong: every required section is present in order (hero, what-it-is with 06/60/7-12 stats, all six exhibits with their correct headline numbers, a rate-it-yourself gross-o-meter, the science explainer with 240 and 9-in-10, parents/teachers, four session times, $14/$18/$52/$9 pricing, the six-field booking form, and full contact block). The dark, big-Archivo-Expanded editorial system is coherent, well-spaced and consistently executed with tasteful cyan accents, and the gross-o-meter panel looks like a real interactive control. The main weakness is impact and audience fit: the

OpenAI: All mandated sections, the six exhibits, gross-o-meter UI, adult assurances, ticket details, school booking form and contact information are visible with the fixed facts largely intact. The black-and-cyan lab aesthetic is coherent and polished, but the very small technical typography and severe tone feel more adult/forensic than funny child-facing museum science. Overall it is a finished, distinctive build, though it stops short of maximum impact because it lacks warmth, playfulness and a stronger general ticket-booking push.

Full run
Cass Werner (kimi-k3)
default reasoning
Cass Werner (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 68.8%, judge 3.8/5
Composite 68.8%Judge 3.8/5
Judge panelAnthropic 3.8/5
single-judge (Claude) 3.8/5 → leave-one-family-out 3.8/5

Anthropic: Every required section is present and content-complete: hero, what-it-is, all six exhibits with facts and ratings, a working gross-o-meter with per-exhibit sliders and Lab-rating comparison, the science explainer, a parents/teachers block covering safety/curriculum/accessibility, the four session times and $14/$18/$52 pricing, the six-field booking form, and visit/contact details - so brief fidelity is excellent. Execution is clean and consistent with aligned cards, tidy type and a coherent mint-green palette, giving it solid craft. The weakness is that the design is very safe and monotone: an

Full run
Yuki Tanabe (deepseek-v4-pro)
default reasoning
Yuki Tanabe (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 56.3%, judge 3.3/5
Composite 56.3%Judge 3.3/5
Judge panelAnthropic 3.0/5OpenAI 3.5/5
single-judge (Claude) 3.0/5 → leave-one-family-out 3.3/5

Anthropic: Content fidelity is high: all sections are present, including per-exhibit gross-o-meter controls with the required named levels and Lab-rating comparisons, correct prices/session times, curriculum codes, accessibility notes, the six-field booking form, and contact details. But the design is a cold, near-monochrome dark editorial treatment with a single cyan accent, which flatly ignores the "make them laugh, then make them learn" brief for a 7-12 kids' exhibition; it feels like an agency/fintech site, not a fun gross-science one. Execution is clean and consistent but monotone, with large empty

OpenAI: Most required content sections are visible and the six exhibits, adult notes, ticketing, form, and contact details are largely present, but the gross-o-meter is underdeveloped in the screenshot: Lab ratings and comparison are not visible, and the booking officer/response note appears absent. The dark cinematic treatment is coherent and polished, but the tiny low-contrast text, vast empty spacing, and very serious tone feel less suited to 7–12 year olds and lack the brief’s laugh-then-learn energy.

Full run
Sol Kepler (kimi-k3)
default reasoning
Sol Kepler (kimi-k3) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 56.3%, judge 3.3/5
Composite 56.3%Judge 3.3/5
Judge panelAnthropic 3.3/5
single-judge (Claude) 3.3/5 → leave-one-family-out 3.3/5

Anthropic: All the required sections are present and content-complete: the six exhibits each carry a name, kid explanation, a big-number surprising fact and a rating; there is a gross-o-meter comparison table, the science explainer (240 students, 9 in 10), parents/teachers safety-curriculum-accessibility copy, session times with $14/$18/$52 pricing, a booking form, and full contact details. The execution is clean and consistent - hairline dividers, aligned columns, tidy typography and a restrained sage-green accent - so craft is solid. But the treatment reads like an austere corporate annual report rathe

Full run
Cass Werner (deepseek-v4-pro)
default reasoning
Cass Werner (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 56.3%, judge 3.3/5
Composite 56.3%Judge 3.3/5
Judge panelAnthropic 3.3/5
single-judge (Claude) 3.3/5 → leave-one-family-out 3.3/5

Anthropic: Content coverage is comprehensive and correct: hero, what-it-is, all six exhibits with kid facts, surprising numbers, adult notes and lab ratings, the gross-o-meter using all five named levels, the science explainer with the 240/9-in-10 stats, parents/teachers (safety, curriculum codes, accessibility), sessions, prices, the full booking form and visit details are all present. Visually it is coherent but generic - white background, single orange accent, plain card grids and default type with heavy whitespace, no illustration or playful energy to match a gross-biology kids' exhibition. Craft is

Full run
Ingrid Kessler (deepseek-v4-pro)
default reasoning
Ingrid Kessler (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 50.0%, judge 3.0/5
Composite 50.0%Judge 3.0/5
Judge panelAnthropic 3.3/5OpenAI 2.8/5
single-judge (Claude) 3.3/5 → leave-one-family-out 3.0/5

Anthropic: All required sections are present and content-complete: hero, what-it-is, all six exhibits with kid facts, numbers and Lab ratings, per-exhibit rate-the-grossness controls showing the named levels (e.g. "Lab: Proper Gross 3/5"), real science, parents/teachers split into safety, curriculum and accessibility, session times, ticket prices, the six-field school booking form, and visit/contact. Execution is a clean, restrained editorial layout on warm cream with a single orange accent, consistent typography and generous spacing, so craft is solid with no broken elements. The weakness is fit-to-audi

OpenAI: Most required sections and factual content are visible, including all six exhibits, adult notes, tickets, school form, and contact details, though the gross-o-meter is very minimal and the named five-level scale is not fully visible. The execution is clean and readable but extremely plain for a children’s gross-science exhibition, with little visual storytelling, playfulness, imagery, or distinctive identity. Craft is consistent, but the page feels more like a sparse text document than a polished museum companion site.

Full run
Elias Byrne (deepseek-v4-pro)
default reasoning
Elias Byrne (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 50.0%, judge 3.0/5
Composite 50.0%Judge 3.0/5
Judge panelAnthropic 3.3/5OpenAI 2.8/5
single-judge (Claude) 3.3/5 → leave-one-family-out 3.0/5

Anthropic: All required sections and fixed content are present and accurate - six exhibits with lab ratings, the gross-o-meter with the five named levels and YOU/LAB comparison, curriculum codes, correct prices, session times, full booking form, and contact details. Execution is clean and internally consistent (tidy alignment, coherent editorial typography, no broken elements), so craft holds up. But the entirely monochrome, imagery-free broadsheet treatment reads as a dry academic programme and completely ignores the primary child audience the brief centres on, giving it no visual hook, warmth, or disti

OpenAI: The page includes the major required sections, correct dates, prices, ratings and contact information, but key exhibit explanations and adult notes are hidden behind collapsed “read more” controls in the visible state, and the gross-o-meter scale is not clearly presented with all named levels. The monochrome editorial treatment is coherent and restrained, but it feels too austere and adult for a funny children’s science exhibition, with small type, sparse controls, and little visual excitement. Overall it is functional and mostly complete, but not especially memorable or polished as a premium

Full run
Mara Vieira (deepseek-v4-pro)
default reasoning
Mara Vieira (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 43.8%, judge 2.8/5
Composite 43.8%Judge 2.8/5
Judge panelAnthropic 2.5/5OpenAI 3.0/5
single-judge (Claude) 2.5/5 → leave-one-family-out 2.8/5

Anthropic: The build covers the brief substantially: hero, what-it-is, all six exhibits with facts and lab ratings, a named five-level gross-o-meter, the real-science quote, parents/teachers safety-curriculum-accessibility, sessions, tickets, the six-field booking form, and full visit/contact details are all present in Australian English. But visually it reads as a near-unstyled document: plain black-on-white system type, hairline-bordered cards, cramped tiny body copy, and two arbitrary floating shapes (a red rectangle, a yellow circle) that feel like placeholders rather than a design point of view. The

OpenAI: The page visibly includes nearly all required sections and fixed facts, including all six exhibits, ratings, science, adult notes, tickets, booking form fields, and visit details, though some required booking reassurance such as Marcus replying within 2 business days is not visible and interactivity cannot be verified from the screenshot. The stark black-and-white brutalist direction is coherent but feels sparse and not especially kid-enticing; there are also craft issues such as the clipped slogan at the left edge and placeholder-like geometric graphics. Overall it is functional and mostly co

Full run
Noor El-Amin (deepseek-v4-pro)
default reasoning
Noor El-Amin (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 40.6%, judge 2.6/5
Composite 40.6%Judge 2.6/5
Judge panelAnthropic 2.8/5OpenAI 2.5/5
single-judge (Claude) 2.8/5 → leave-one-family-out 2.6/5

Anthropic: Content coverage is strong: all six exhibits with kid facts, numbers, adult notes and lab-rating dots, plus the explainer, safety/curriculum/accessibility, sessions, prices, booking form and contact details are all present, and a gross-o-meter block with named exhibit pills appears. But the visual design is a stark black-and-white editorial layout that badly misreads the brief's brief - a laugh-then-learn gross-biology show for 7-to-12 year olds should feel playful and colourful, and this reads as an austere newspaper with vast empty gutters and no illustration or energy. Craft is clean but sp

OpenAI: Most required sections and core facts are visible, including the six exhibits, adult/safety material, tickets, school form, and contact details, but the gross-o-meter appears incomplete in the screenshot with no visible five-level scale or Lab comparison, and some fixed details are missing or overly abbreviated. The stark black-and-white editorial style is coherent but underwhelming for a children’s gross science exhibition, and visible unfinished elements like the literal map placeholder and awkward empty spacing reduce craft and impact.

Full run
Sol Kepler (deepseek-v4-pro)
default reasoning
Sol Kepler (deepseek-v4-pro) rendering of the The Gross Lab children's museum exhibition companion site benchmark - composite 37.5%, judge 2.5/5
Composite 37.5%Judge 2.5/5
Judge panelAnthropic 2.5/5
single-judge (Claude) 2.5/5 → leave-one-family-out 2.5/5

Anthropic: Content coverage is strong: all six exhibits with kid facts, surprising numbers, adult notes and Lab ratings are present, the gross-o-meter appears with the five named levels, and the real-science, parents/teachers, sessions/prices, booking form and contact sections are all there. But visually it is an almost entirely unstyled black-on-white document - no colour, no imagery, no theming, flat type hierarchy and thin generic borders - which utterly fails the "make them laugh, then make them learn" brief for a kids' exhibition and reads as a wireframe. It is tidy and consistently aligned with not

Full run