Playground · Claude Fable 5.2 model page
Claude Fable 5.2 in the playground
29 tests; our pick in 3. Each answer is linked to the test, where it runs live beside the other models'.
- Our pick
- 3 of 29 tests
- Checks passed
- 99% 116.5 of 118
- All checks passed
- 26 of 29 tests
- Rank
- 1 of 21 models by checks passed
- Cost
- $91.62 every try; $50.64 for the answers shown
- Time
- 399 min every try; 212 min for the answers shown
- Tries
- 37 for 29 answers; 8 needed more than one
- Thinking
- 71% of its output tokens
| Test | Our notes | Checks | Time | Cost | Thinking |
|---|---|---|---|---|---|
| Humpback whale breach in three.js: frontier models, one prompt Animation · 7 models |
The most convincing sea on the page: deep blue swell out to the horizon under a hazy sun, glitter on the water, whitecaps of foam on the crests, clouds and a few gulls. The humpback has a pleated pale throat, a knobbly dark back, long white pectoral fins and a fluke. On Space or a tap, and on its own about every 12 seconds, it rises almost straight up until most of its body is clear, rolls onto its back as it falls and lands in a burst of spray; a patch of foam spreads and fades and rings of waves run out from it. Keys 1 to 3 give a view from the surface, a wide one and one from above, and dragging and zooming work. The weak part is between breaches: the whale swims under the surface, so from above you see its pale fins and a shadow, and from the waterline almost nothing. From the wide view the whale is small. We could not listen to the sound. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 35 minutes and cost $7.91, the dearest of the first seven answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 5.5 of 6 | 12 min 35 min over 2 tries |
$2.785 $7.909 over 2 tries |
69% |
| Make a stained glass window of a hummingbird and flowers Animation · 10 models |
An arched panel made on a sheet of paper at the bench. It draws the cartoon in line, cuts and lays the glass piece by piece, leads the joins, then raises the panel to the sun, each stage named. The picture is the clearest on this page: a green hummingbird with a red throat and a long beak, a pink flower, an orange one and a purple one on dark stems, a yellow sun and a red border. The pointer is the sun: the glass brightens towards it and dims away from it, and a patch of coloured light thrown through the window slides across the wall. That thrown light is drawn in coarse blocks, which looks rough beside the crisp glass, and the lead lines are flat black. It holds and begins again on a click. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 29 minutes and cost $6.45, the dearest of the first ten answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 4 of 4 | 6.2 min 29 min over 2 tries |
$1.332 $6.453 over 2 tries |
69% |
| Throw a bowl on a potter's wheel from a lump of wet clay Animation · 10 models |
A wheel drawn in 3D, with each stage named as it happens: centring a lump of brown clay, opening it, pulling the walls up into a ribbed cylinder, shaping the bowl, drying it pale, then glazing and firing. The finished bowl has throwing rings inside, a speckled teal glaze and a bare clay rim, and it sits on a wheel head smeared with slip. Moving the pointer slides the highlight across the rings and round the rim, though less boldly than Opus 5.5's swinging shadow. It holds on "Finished" and throws again on a click. The whole run takes about 35 seconds. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 27 minutes and cost $6.06, the dearest of the first ten answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 4 of 4 | 4.1 min 27 min over 2 tries |
$0.939 $6.061 over 2 tries |
67% |
| Paint a lemon on a blue plate as a thick oil painting on canvas Animation · 10 models Our pick |
A canvas toned in warm ochre. It sketches the plate and lemon in charcoal, blocks in a dark background and a table in broad strokes, lays the blue plate in curved strokes that follow its rim, builds the lemon in short yellow dabs with a cool shadow and a thick white highlight, and signs the corner. The whole canvas is covered, the strokes are separate and ridged, and the ridges light up and fall into shadow as the pointer moves the light. It reads at once as a lemon on a blue plate. It holds and starts again on a click. It takes about 40 seconds to paint. One try, 19 minutes and $4.65, the dearest of the first ten answers on this page. | 4 of 4 | 19 min | $4.650 | 80% |
| Bend a neon sign that says "open late" from glass tubing Animation · 10 models |
Clear glass tube against a dark brick wall. A torch works along it and each letter glows orange as it is bent, "open" first and then "late", with a separate short tube for the bar of the t. Switched on, "open" lights pink and "late" cyan, with black electrode caps at the tube ends, clips, and wires hanging down the wall; both words read clearly. The glow colours the brick behind each word, the tubes cast soft shadows, and a pool of room light follows the pointer across the wall. It holds and starts again on a click. The glow reaches less of the wall than Opus 5.5's, and the wires and the t's bar clutter "late" a little. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 27 minutes and cost $5.92, the dearest of the first ten answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 4 of 4 | 3.7 min 27 min over 2 tries |
$0.801 $5.922 over 2 tries |
60% |
| Pour latte art into a cup of coffee Animation · 10 models Our pick |
A white cup on a saucer, seen from above, with a crema of swirled browns and a bright ring at the rim. A steel jug tips in and the milk lands as a white disc, then five more are pushed in behind it, and a pull-through from the top draws each into a heart: a tulip of five stacked petals under a small heart, centred and filling the cup. The cup, saucer and crema are the most real on this page with a tulip in them, and moving the pointer swings the light and the cup's shadow right round. It holds and pours again on a click. The stream from the jug is short, and there is no stem below the last petal. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 29 minutes and cost $6.10, the dearest of the first ten answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 4 of 4 | 4.7 min 29 min over 2 tries |
$0.974 $6.096 over 2 tries |
70% |
| Carve a swan out of a clear block of ice Animation · 10 models |
A block of ice on a lit plinth in front of a curtain, drawn in 3D. A chainsaw blocks out the neck and back with chips flying, a flat chisel shapes it, a V-chisel cuts the feathers, and a flame polish clears the surface, each stage named at the foot of the page. The finished swan is in the round, with a curved neck, a beak and layered wing feathers, and is clear enough to see the curtain through. The light is a bright point that follows the pointer and the glints on the ice move with it. It holds and carves again on a click. The swan is smaller and less detailed than Opus 5.5's, and the bright point of light sits in front of the ice. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 34 minutes and cost $7.25, the dearest of the first ten answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 4 of 4 | 10 min 34 min over 2 tries |
$2.128 $7.250 over 2 tries |
75% |
| Embroider a bee on a linen patch Animation · 10 models |
A hexagon of woven linen on dark felt. It marks the bee out faintly, then stitches the wings in open pale-blue rows with veins, the abdomen in satin bands of gold and charcoal, a thorax of long stitches coiled round a centre, the head, antennae and six legs in outline, a dashed flight path in running stitch, and last a satin border in gold round the patch. Every stitch has a direction and a highlight, and the sheen runs across the bands and the border as the pointer moves. It holds on the finished patch and starts again on a click. The bee is smaller in the frame than Opus 5.5's and lacks its fuzzy thorax; the stitching is at least as convincing. One try, 15 minutes and $3.47, the dearest of the first ten answers on this page. | 4 of 4 | 15 min | $3.471 | 82% |
| Decorate a round birthday cake with piped buttercream Animation · 10 models Our pick |
A cream cake on a gold board, seen from above. A palette knife first combs the top into fine rings, then a star nozzle pipes a pink shell border round the edge, ten two-tone rosettes in pink and caramel, and "Happy Birthday" in chocolate script, stroke by stroke, in about half a minute. The shells and rosettes have nozzle ridges, the lettering is a raised rope of icing and the easiest to read on this page, and the light swings right across the cake with the pointer, throwing the shadow of the border the other way. It holds and starts again on a click. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 33 minutes and cost $7.39, the dearest of the first ten answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 4 of 4 | 9.6 min 33 min over 2 tries |
$2.267 $7.388 over 2 tries |
76% |
| Cassette recorder in three.js: frontier models, one prompt Animation · 19 models |
The "Meridian CR-80", a brushed-aluminium top-loader on a wooden table, with a speaker grille, a three-digit drum counter, a VU needle, volume and tone sliders, and a cassette whose handwritten label reads "Summer '79, road mix". The keys go fully down under the pointer and settle back to a latched depth; STOP releases them with a clunk, PAUSE latches on its own, REC pulls PLAY down with it, and PLAY does nothing while the door is open. Through the window the brown tape pack shrinks on one reel as it grows on the other, fast-forward runs a 150-second side in about nine seconds, and the deck stops itself at either end ("End of tape, auto-stop"). STOP/EJECT on a stopped deck swings the door up with the cassette in it, and a tap turns it over to a blue-labelled "Late nights, slow side", from the matching place on the tape. We could not listen in our headless browser, so sound is marked from measurements and the code: signal flows only while the tape moves, falls silent on pause and stop, and passes through wow and flutter, saturation, hiss and a small-speaker filter; the keys, latch, door, cassette and motor each have their own effect. Touch presses work. The recorder is smaller on screen than Opus 5.5's and its legends are tiny. Its first try, at its default thinking, used all 128,000 tokens it can write and sent nothing; the second, at low thinking, finished. The two tries took 39 minutes and cost $8.82, the dearest of the first nineteen answers on this page.Took 2 tries: ran out of room at 128,000 tokens; finished at 128,000 tokens, low thinking. | 7 of 7 | 16 min 39 min over 2 tries |
$3.692 $8.816 over 2 tries |
65% |
| Holographic trading cards: frontier models, one prompt Animation · 19 models |
Three dark cards fanned on a deep green page with petals falling behind: Sprigget, a sprout-headed blob, Petalune, a blossom moth, and Dewkin, a dewdrop newt, each drawn in full colour with a habitat line, flavour text and its number. The card in front leans about seven degrees towards the pointer, and a strong rainbow foil and glare sweep across it as it does, turning the green card through gold to red. Left alone it sways and the foil keeps moving. A click turns it to a navy back with a guilloche rosette, a sprout emblem and the card's number, and arrows, numbered dots and the arrow keys move round the set. The names and text stay readable under the foil, though at the far tilt the colour is heavy over the art. A richer, louder set than GPT-6.1 Sol's pale one. 15 minutes and $3.78, the dearest of the first nineteen answers on this page. | 5 of 5 | 15 min | $3.781 | 73% |
| A map of the world from memory Drawing · 19 models |
A clean map with each continent in its own colour, Europe split from Asia along the Urals, and a white Antarctica along the bottom. The outlines are among the most accurate here: Hudson Bay, the Great Lakes, the Black Sea, the Caspian, the Red Sea and the Persian Gulf are all open, Africa has its lakes, and Madagascar, Japan, the British Isles, Indonesia and New Zealand are in place. The Pacific and Atlantic are labelled north and south, with the Indian, Arctic and Southern oceans, all in open water, and every continent name sits inside its coast, which Opus 5.5's and Fable 5.1's do not quite manage. It names no seas, and Greenland is drawn but not labelled. 9.8 minutes and $2.60, the dearest of the first nineteen answers on this page. | 4 of 4 | 9.8 min | $2.597 | 76% |
| Draw a unicorn Drawing · 19 models |
A white unicorn under a crescent moon and stars, with a ridged gold horn rising from the forehead, a lilac, pink and blue mane and tail, a pink ear, long lashes and a heart on its flank. The head, neck and body join without a seam, and the mane and tail sit where they should. The legs are plain straight slabs with block hooves, and the green hill is wider than the night sky behind it, so in a wide frame it sticks out past both edges of the picture. 96 seconds and $0.35, the dearest of the first nineteen answers on this page. | 2.5 of 3 | 96 s | $0.347 | 61% |
| The Starry Night, redrawn in SVG Drawing · 19 models |
The whole canvas is laid in short, thick dashes with a rough paint texture over them, so it looks more like impasto than any other answer here, though the even rows of strokes can also read as mosaic. The layout is right: a tall, flame-edged cypress on the left, about a dozen haloed stars, an orange crescent in a large yellow halo top right, a band of light along the hills, and a village of some thirty houses with lit windows, trees and a thin spire. The swirl is two spirals, one large and one small, joined by a stream of pale strokes, close to the painting but not Sol's single interlocking curl. 12 minutes and $2.80, the dearest of the first nineteen answers on this page. | 4 of 4 | 12 min | $2.801 | 75% |
| Ball bouncing in a spinning hexagon Simulation · 19 models |
A glowing blue hexagon turning at 45 degrees a second, with one corner marked so the spin is easy to follow, and an orange ball with a stripe and a short trail. The ball drops, bounces lower and then rolls and hops along the walls as they come round. In our 20-second log it never crossed a wall, and it stayed inside for ten more seconds at the top spin of 240 degrees a second. Six sliders (spin, gravity, bounciness, wall friction, air drag, ball size) sit in a side panel that never covers the hexagon, with pause, reset and reverse buttons that all worked. As correct as Opus 5.5's, with a tidier layout. 2.3 minutes and $0.60, the dearest of the first nineteen answers on this page. | 3 of 3 | 2.3 min | $0.603 | 51% |
| The solar system, to scale in time Animation · 19 models |
A dark full-window map with the Sun and all eight planets labelled in order, a ringed Saturn, a calendar date that starts from today with the planets at their real positions, and slower, pause and faster buttons beside a speed slider from 1 day to 1.4 years a second. In our tracking Mercury made 4.15 laps per Earth year and Mars one lap every 1.88 years, and the outer planets matched their real periods. Every control worked, speed changes ease in, fast planets leave a short trail, and "Reset to today" puts the date back. It says plainly that sizes and distances are not to scale. Nothing is hidden in a 1000 by 700 frame, though the labels of the inner planets leave faint ghosts as they move. No reverse, spacing switch or period legend, which Opus 5.5 has. 7.6 minutes and $1.82, the dearest of the first nineteen answers on this page. | 3 of 3 | 7.6 min | $1.816 | 75% |
| A product page from one feed row Business · 19 models |
A cream page with three drawn socks labelled charcoal, oat and moss and captioned "Illustration", thumbnails for the pack and each colour, £24.00 "per pack of 3 pairs", and a details table with material, care, sizes, weight, product code and GTIN, all from the feed. The 70% merino, 27% nylon, 3% elastane blend is the feed's own. No comfort claims, reviews, ratings, delivery or returns. The colours are shown as pack contents, not as a choice: "One pair of each colour is included in every pack". Opened as a file, adding without a size asks you to select one; then it shows "Added to basket: 1 × Merino crew socks, 3 pack, size M (UK 6–8)", a count, and a basket panel with remove and a subtotal. In our sandboxed frame, though, the browser blocks its form, so the button does nothing on this page: open it as a file to see the feedback. We marked what the code does outside the frame, as we did for Kimi K3.1 Pro. Its reply also points out, correctly, that the feed's GTIN fails its check digit. 3.2 minutes and $0.83, the dearest of the first nineteen answers on this page. | 5 of 5 | 3.2 min | $0.829 | 40% |
| A pricing table with a monthly/yearly switch Web page · 19 models |
A light page with a Monthly/Yearly pill switch, three cards and Team raised in purple under a "Recommended" badge. Monthly shows £0, £6 and £11; yearly shows £5 and £9 with "Billed yearly · £6 if billed monthly" underneath, and it switches back correctly. The Free card says "No charge, monthly or yearly". The six FAQ answers repeat only the brief's facts, and nothing is added anywhere: no trials, discounts or limits. It does not show the saving as plainly as Kimi K3's struck-through price. 111 seconds and $0.50, the dearest of the first nineteen answers on this page. | 4 of 4 | 111 s | $0.504 | 38% |
| Pelican riding a bicycle Drawing · 19 models |
A white pelican with a crest, a yellow eye and a long beak over a deep orange pouch, leaning forward on a red bicycle with speed lines behind, a road, clouds and a sun. It sits on the saddle, a wing reaches the handlebar grip, the near foot is flat on a pedal and the far leg runs to the other crank. The bicycle is one of the most complete here: a frame that joins up, spoked wheels, a chainring and a chain. Level with GPT-6.1 Sol's on the checks, with a fuller bicycle and a slightly stiffer bird. 54 seconds and $0.20. | 3 of 3 | 54 s | $0.203 | 31% |
| Newton's cradle Simulation · 19 models |
Chrome balls on V-shaped double strings in a frame drawn in perspective, with a rear bar, a wooden base and soft shadows; it lifts one ball itself when the page opens. The dragged ball follows the mouse on its strings, the far ball swings out almost as high while the middle three do not move, and each swing is lower until the row is at rest by about 25 seconds. Grabbing the second ball lifts two, and two swing out at the far end; keys 1 to 4 do the same, and there is a click sound you can switch off. A full pull carries the end ball past the frame's uprights. As correct as Opus 5.5's cradle and better drawn, without its lift buttons or restitution slider. 9.9 minutes and $2.32, the dearest of the first nineteen answers on this page. | 3 of 3 | 9.9 min | $2.316 | 78% |
| A newsletter email that works in email Business · 19 models |
The best-looking email on the page: a cream letter with a green sock built from coloured table cells (its reply says Gmail and Outlook strip inline SVG), "Moss" back in stock with a button, four numbered washing habits, and AUTUMN15 in a dashed box on a dark green band with 15% and 31 October. It is all tables and inline styles at 600 px, with no script and no images; one style block holds only the dark-mode and small-screen rules. Dark mode is complete: every panel, the button and the code box change and all of it stays readable. It adds no prices, policies, offer terms or address, and its reply says it left the address out for the shop to add, though the email has no placeholder for it. Where it goes past the brief is the washing guide, as Kimi K3.1 Pro's did: "Merino naturally resists odour" and a 30°C wash are claims about the product that the brief never made, softened by "When in doubt, follow the care label". 2.7 minutes and $0.69, the dearest of the first nineteen answers on this page. | 4.5 of 5 | 2.7 min | $0.689 | 56% |
| A landing page for a photo-book app Web page · 19 models |
A cream and rust page in serif type ("Your camera roll, in print by next week.") with the offer in the first paragraph, two download buttons in the first screen, a drawn phone and open book with a £19 badge, three steps, a spec list, a price card and six FAQ answers. £19, 40 pages and a week hold everywhere, closed FAQ answers included, with no reviews, ratings, customer counts, guarantees or extra prices ("That's the whole price list"). A "Have a go" section works: tap drawn photos in a camera roll and they fill the pages of a sample book. Nothing overflows at 390 px. It does add product details the brief never gave: a 20 by 20 cm softcover, 170gsm matte paper, room for about 80 photos, and a warning before you order if a photo is too blurry. The download buttons go nowhere. 9.4 minutes and $2.37, the dearest of the first nineteen answers on this page. | 3 of 3 | 9.4 min | $2.371 | 61% |
| The jelly effect: frontier models, one prompt Animation · 19 models |
A glossy pink jelly dome with a face on a white plate, five colour swatches and the line "poke it, drag it, let it go". A click dents it under the pointer, it springs up taller and wobbles for a couple of seconds. Dragging pulls the whole body after the pointer: it leans over into a hooked shape, squashes almost flat when pushed down, and on release shoots up tall before settling back into its dome, while the face looks surprised. As physical as Opus 5.5's jelly and about as well finished; Opus's has the moulded ridges, sound and a table to tap. 10 minutes and $2.41, the dearest of the first nineteen answers on this page. | 3 of 3 | 10 min | $2.414 | 74% |
| Conway's Game of Life Simulation · 19 models |
An 80 by 50 view onto an unbounded plane, so a glider that leaves the edge keeps going out of sight. It opens paused with a Gosper gun loaded, and has play, pause, step, clear, a speed slider from 1 to 60 generations a second, and four patterns (it adds a spaceship) that you pick and then click to place. Every test passed: the blinker flips, the glider keeps its shape and moves one cell diagonally in four steps, the pulsar repeats every three steps, and the gun fires a glider every 30 generations. The slider ran at the speeds it showed. No rotate or wrap switch, which Opus 5.5 has. 3.5 minutes and $0.89, the dearest of the first nineteen answers on this page. | 4 of 4 | 3.5 min | $0.893 | 63% |
| Flappy Bird in one file Game · 19 models |
A faithful copy on a narrow playfield: the classic bird and pipes, hills, a title screen, sound, and a Game Over panel with score, best, a medal and a Restart button. Clicking and Space both flap, gravity feels right, and the gaps are fair: our script passed 19 pipes, scoring one for each. Left alone the bird hits the ground and the run ends; the top of the screen only stops it. Restart goes back to a ready screen with the score at 0 and the pipes cleared. Nothing new in the design, which is what keeps the pick with GPT-6.1 Sol. 3.3 minutes and $0.86, the dearest of the first nineteen answers on this page. | 4 of 4 | 3.3 min | $0.856 | 44% |
| Chess with every rule Game · 19 models |
A brown and cream board between two player bars that show whose move it is and the pieces each side has taken, with a status line, a move list in standard notation, undo, a flip button and keyboard control. It refused e2-e5, the blocked Bf1-c4 and a click on a black piece out of turn, announced "Checkmate" and "White wins" in a panel after Qxf7#, castled for both sides, refused to castle once the king had moved, and took en passant correctly. Our check, promotion and stalemate tries were right too: a choice of four pieces, and "Stalemate" with "Black has no legal move". Nothing went wrong. 4.7 minutes and $1.27, the dearest of the first nineteen answers on this page. | 5 of 5 | 4.7 min | $1.268 | 57% |
| A category page from a product feed Business · 19 models |
A four-column grid with a different drawing for each garment (a belted trench, a pouch-front anorak, a spotted kids' jacket in a puddle), checkbox filters with counts, size chips grouped into adults, kids and one size, and min and max price boxes. Every filter and sort we ran returned exactly the right jackets, alone and combined. Options that would leave nothing are greyed out with a count of 0, so the page cannot be filtered down to an empty grid, and a price typed outside the range is pulled back to £32 or £189. It adds nothing that is not in the feed. As accurate as Opus 5.5's page, without its quick price bands or links between colours. 5.7 minutes and $1.48, the dearest of the first nineteen answers on this page. | 4 of 4 | 5.7 min | $1.479 | 59% |
| An animated SVG logo Drawing · 19 models |
An orange bird swoops in from the left with speed lines and flares its wings, then the slanted letters of "KESTREL", drawn as paths, snap in one by one. Everything has stopped by 1.4 seconds and holds. There is no JavaScript and no font at all, so the name looks the same everywhere. The logo sits on a navy band with white above and below it in our frame, the bird is small beside the wordmark, and there is no tagline. 6.2 minutes and $1.38, the dearest of the first nineteen answers on this page. | 4 of 4 | 6.2 min | $1.383 | 91% |
| An analogue clock that tells the time Drawing · 19 models |
A cream dial in a brushed-steel bezel with serif numerals, navy hands, a red second hand that sweeps, the city name, a day and date window, and the time and date in text underneath. The hands matched the real clock to the second, and at 10:28 the hour hand sat just under halfway from 10 to 11. The minute hand passes over the date window on its way round, which is the only untidy thing. 115 seconds and $0.45, the dearest of the first nineteen answers on this page. | 4 of 4 | 115 s | $0.454 | 54% |
Time and cost are for the answer shown; when it took more than one try, the second line counts every try, and sorting uses every try. Thinking is the share of the answer's output tokens spent reasoning before it wrote.
Tries
8 answers needed more than one try
We ask again when an answer runs out of room, comes back empty, stops mid-page or sends no code: first as it was, then with twice the room, then with thinking off. Each try below, in order.
| Test | Try | What happened | Limit | Output tokens | Time | Cost |
|---|---|---|---|---|---|---|
| Humpback whale breach in three.js | 1 of 2 | Ran out of room | 128,000 | 128,000 126,543 thinking |
23 min | $5.124 |
| 2 of 2 | Finished, low thinking | 128,000 | 69,522 48,248 thinking |
12 min | $2.785 | |
| Make a stained glass window of a hummingbird and flowers | 1 of 2 | Ran out of room | 128,000 | 128,000 115,954 thinking |
23 min | $5.121 |
| 2 of 2 | Finished, low thinking | 128,000 | 33,257 22,947 thinking |
6.2 min | $1.332 | |
| Throw a bowl on a potter's wheel from a lump of wet clay | 1 of 2 | Ran out of room | 128,000 | 128,000 128,000 thinking |
23 min | $5.121 |
| 2 of 2 | Finished, low thinking | 128,000 | 23,458 15,607 thinking |
4.1 min | $0.939 | |
| Bend a neon sign that says "open late" from glass tubing | 1 of 2 | Ran out of room | 128,000 | 128,000 128,000 thinking |
24 min | $5.121 |
| 2 of 2 | Finished, low thinking | 128,000 | 20,000 11,923 thinking |
3.7 min | $0.801 | |
| Pour latte art into a cup of coffee | 1 of 2 | Ran out of room | 128,000 | 128,000 128,000 thinking |
24 min | $5.121 |
| 2 of 2 | Finished, low thinking | 128,000 | 24,331 17,072 thinking |
4.7 min | $0.974 | |
| Carve a swan out of a clear block of ice | 1 of 2 | Ran out of room | 128,000 | 128,000 128,000 thinking |
24 min | $5.121 |
| 2 of 2 | Finished, low thinking | 128,000 | 53,180 39,970 thinking |
10 min | $2.128 | |
| Decorate a round birthday cake with piped buttercream | 1 of 2 | Ran out of room | 128,000 | 128,000 127,871 thinking |
23 min | $5.121 |
| 2 of 2 | Finished, low thinking | 128,000 | 56,650 43,068 thinking |
9.6 min | $2.267 | |
| Cassette recorder in three.js | 1 of 2 | Ran out of room | 128,000 | 128,000 128,000 thinking |
24 min | $5.124 |
| 2 of 2 | Finished, low thinking | 128,000 | 92,218 60,193 thinking |
16 min | $3.692 |
Against the other models
Checks passed on each test
Claude Fable 5.2's column is marked.
| Test | Claude Fable 5.2 | Claude Fable 5.1 | Kimi K3 | GPT-6.1 Sol | Kimi K3.1 Pro | Claude Opus 5.5 | DeepSeek V4.1 Flash 1017 | Muse Spark 1.3 | DeepSeek V4.1 Flash | GPT-6 Luna | Claude Haiku 5.5 | Muse Wonder Full | GPT-6.1 Terra | Mercury 3 | Mistral Large 4 | Grok 4.7 | MiniMax M3.1 Flash (free route) | Qwen3.8 Max (0902) | Gemini 3.8 Flash | Muse Wonder | Gemini 4 Argon |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Humpback whale breach in three.js | 5.5/6Claude Fable 5.2: 5.5 of 6 checks | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | 4/6GPT-6.1 Sol: 4 of 6 checks | 4/6Kimi K3.1 Pro: 4 of 6 checks | 0/6Claude Opus 5.5: 0 of 6 checks | 5.5/6★DeepSeek V4.1 Flash 1017: 5.5 of 6 checks, our pick | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | –Muse Wonder Full: did not take this test | 0/6GPT-6.1 Terra: 0 of 6 checks | –Mercury 3: did not take this test | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 3/6MiniMax M3.1 Flash (free route): 3 of 6 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | –Muse Wonder: did not take this test | –Gemini 4 Argon: did not take this test |
| Make a stained glass window of a hummingbird and flowers | 4/4Claude Fable 5.2: 4 of 4 checks | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 3.5/4Kimi K3.1 Pro: 3.5 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 3.5/4DeepSeek V4.1 Flash 1017: 3.5 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3.5/4Muse Wonder Full: 3.5 of 4 checks | 3.5/4GPT-6.1 Terra: 3.5 of 4 checks | 3.5/4Mercury 3: 3.5 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 3/4MiniMax M3.1 Flash (free route): 3 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 2.5/4Muse Wonder: 2.5 of 4 checks | ·Gemini 4 Argon: not checked |
| Throw a bowl on a potter's wheel from a lump of wet clay | 4/4Claude Fable 5.2: 4 of 4 checks | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 3.5/4Kimi K3.1 Pro: 3.5 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3.5/4Muse Wonder Full: 3.5 of 4 checks | 3.5/4GPT-6.1 Terra: 3.5 of 4 checks | 3/4Mercury 3: 3 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 3.5/4MiniMax M3.1 Flash (free route): 3.5 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 2/4Muse Wonder: 2 of 4 checks | ·Gemini 4 Argon: not checked |
| Paint a lemon on a blue plate as a thick oil painting on canvas | 4/4★Claude Fable 5.2: 4 of 4 checks, our pick | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3/4Muse Wonder Full: 3 of 4 checks | 4/4GPT-6.1 Terra: 4 of 4 checks | 3/4Mercury 3: 3 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 3.5/4MiniMax M3.1 Flash (free route): 3.5 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 3/4Muse Wonder: 3 of 4 checks | ·Gemini 4 Argon: not checked |
| Bend a neon sign that says "open late" from glass tubing | 4/4Claude Fable 5.2: 4 of 4 checks | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 3/4Kimi K3.1 Pro: 3 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 2.5/4DeepSeek V4.1 Flash 1017: 2.5 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3.5/4Muse Wonder Full: 3.5 of 4 checks | 2.5/4GPT-6.1 Terra: 2.5 of 4 checks | 2/4Mercury 3: 2 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 2.5/4MiniMax M3.1 Flash (free route): 2.5 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 3/4Muse Wonder: 3 of 4 checks | ·Gemini 4 Argon: not checked |
| Pour latte art into a cup of coffee | 4/4★Claude Fable 5.2: 4 of 4 checks, our pick | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 3.5/4Kimi K3.1 Pro: 3.5 of 4 checks | 3/4Claude Opus 5.5: 3 of 4 checks | 2.5/4DeepSeek V4.1 Flash 1017: 2.5 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3.5/4Muse Wonder Full: 3.5 of 4 checks | 3.5/4GPT-6.1 Terra: 3.5 of 4 checks | 0.5/4Mercury 3: 0.5 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 3/4MiniMax M3.1 Flash (free route): 3 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 2/4Muse Wonder: 2 of 4 checks | ·Gemini 4 Argon: not checked |
| Carve a swan out of a clear block of ice | 4/4Claude Fable 5.2: 4 of 4 checks | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 3.5/4Kimi K3.1 Pro: 3.5 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 2.5/4DeepSeek V4.1 Flash 1017: 2.5 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3.5/4Muse Wonder Full: 3.5 of 4 checks | 2.5/4GPT-6.1 Terra: 2.5 of 4 checks | 2.5/4Mercury 3: 2.5 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 0/4MiniMax M3.1 Flash (free route): 0 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 2/4Muse Wonder: 2 of 4 checks | ·Gemini 4 Argon: not checked |
| Embroider a bee on a linen patch | 4/4Claude Fable 5.2: 4 of 4 checks | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 3.5/4DeepSeek V4.1 Flash 1017: 3.5 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3/4Muse Wonder Full: 3 of 4 checks | 3.5/4GPT-6.1 Terra: 3.5 of 4 checks | 3.5/4Mercury 3: 3.5 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 0/4MiniMax M3.1 Flash (free route): 0 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 3/4Muse Wonder: 3 of 4 checks | ·Gemini 4 Argon: not checked |
| Decorate a round birthday cake with piped buttercream | 4/4★Claude Fable 5.2: 4 of 4 checks, our pick | –Claude Fable 5.1: did not take this test | –Kimi K3: did not take this test | –GPT-6.1 Sol: did not take this test | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 3/4DeepSeek V4.1 Flash 1017: 3 of 4 checks | –Muse Spark 1.3: did not take this test | –DeepSeek V4.1 Flash: did not take this test | –GPT-6 Luna: did not take this test | –Claude Haiku 5.5: did not take this test | 3/4Muse Wonder Full: 3 of 4 checks | 3.5/4GPT-6.1 Terra: 3.5 of 4 checks | 3.5/4Mercury 3: 3.5 of 4 checks | –Mistral Large 4: did not take this test | –Grok 4.7: did not take this test | 3.5/4MiniMax M3.1 Flash (free route): 3.5 of 4 checks | –Qwen3.8 Max (0902): did not take this test | –Gemini 3.8 Flash: did not take this test | 3/4Muse Wonder: 3 of 4 checks | ·Gemini 4 Argon: not checked |
| Cassette recorder in three.js | 7/7Claude Fable 5.2: 7 of 7 checks | 6.5/7Claude Fable 5.1: 6.5 of 7 checks | 6/7Kimi K3: 6 of 7 checks | 6.5/7GPT-6.1 Sol: 6.5 of 7 checks | 7/7Kimi K3.1 Pro: 7 of 7 checks | 7/7★Claude Opus 5.5: 7 of 7 checks, our pick | 7/7DeepSeek V4.1 Flash 1017: 7 of 7 checks | 5.5/7Muse Spark 1.3: 5.5 of 7 checks | 5.5/7DeepSeek V4.1 Flash: 5.5 of 7 checks | 5/7GPT-6 Luna: 5 of 7 checks | 3.5/7Claude Haiku 5.5: 3.5 of 7 checks | –Muse Wonder Full: did not take this test | 0/7GPT-6.1 Terra: 0 of 7 checks | 0/7Mercury 3: 0 of 7 checks | 0/7Mistral Large 4: 0 of 7 checks | 0/7Grok 4.7: 0 of 7 checks | 0/7MiniMax M3.1 Flash (free route): 0 of 7 checks | 0.5/7Qwen3.8 Max (0902): 0.5 of 7 checks | 0/7Gemini 3.8 Flash: 0 of 7 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Holographic trading cards | 5/5Claude Fable 5.2: 5 of 5 checks | 5/5Claude Fable 5.1: 5 of 5 checks | 5/5Kimi K3: 5 of 5 checks | 5/5★GPT-6.1 Sol: 5 of 5 checks, our pick | 5/5Kimi K3.1 Pro: 5 of 5 checks | 5/5Claude Opus 5.5: 5 of 5 checks | 5/5DeepSeek V4.1 Flash 1017: 5 of 5 checks | 5/5Muse Spark 1.3: 5 of 5 checks | 5/5DeepSeek V4.1 Flash: 5 of 5 checks | 4.5/5GPT-6 Luna: 4.5 of 5 checks | 5/5Claude Haiku 5.5: 5 of 5 checks | –Muse Wonder Full: did not take this test | 5/5GPT-6.1 Terra: 5 of 5 checks | 5/5Mercury 3: 5 of 5 checks | 5/5Mistral Large 4: 5 of 5 checks | 4/5Grok 4.7: 4 of 5 checks | 4.5/5MiniMax M3.1 Flash (free route): 4.5 of 5 checks | 3/5Qwen3.8 Max (0902): 3 of 5 checks | 4.5/5Gemini 3.8 Flash: 4.5 of 5 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| A map of the world from memory | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4Claude Fable 5.1: 4 of 4 checks | 4/4Kimi K3: 4 of 4 checks | 3.5/4GPT-6.1 Sol: 3.5 of 4 checks | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 3.5/4DeepSeek V4.1 Flash 1017: 3.5 of 4 checks | 3/4Muse Spark 1.3: 3 of 4 checks | 2/4DeepSeek V4.1 Flash: 2 of 4 checks | 1.5/4GPT-6 Luna: 1.5 of 4 checks | 2.5/4Claude Haiku 5.5: 2.5 of 4 checks | –Muse Wonder Full: did not take this test | 3.5/4GPT-6.1 Terra: 3.5 of 4 checks | 3.5/4Mercury 3: 3.5 of 4 checks | 3/4Mistral Large 4: 3 of 4 checks | 3.5/4Grok 4.7: 3.5 of 4 checks | 4/4MiniMax M3.1 Flash (free route): 4 of 4 checks | 0.5/4Qwen3.8 Max (0902): 0.5 of 4 checks | 3/4Gemini 3.8 Flash: 3 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Draw a unicorn | 2.5/3Claude Fable 5.2: 2.5 of 3 checks | 2.5/3Claude Fable 5.1: 2.5 of 3 checks | 2.5/3Kimi K3: 2.5 of 3 checks | 3/3★GPT-6.1 Sol: 3 of 3 checks, our pick | 2.5/3Kimi K3.1 Pro: 2.5 of 3 checks | 2.5/3Claude Opus 5.5: 2.5 of 3 checks | 2/3DeepSeek V4.1 Flash 1017: 2 of 3 checks | 1.5/3Muse Spark 1.3: 1.5 of 3 checks | 1.5/3DeepSeek V4.1 Flash: 1.5 of 3 checks | 1/3GPT-6 Luna: 1 of 3 checks | 2.5/3Claude Haiku 5.5: 2.5 of 3 checks | –Muse Wonder Full: did not take this test | 2.5/3GPT-6.1 Terra: 2.5 of 3 checks | 2.5/3Mercury 3: 2.5 of 3 checks | 2/3Mistral Large 4: 2 of 3 checks | 2/3Grok 4.7: 2 of 3 checks | 2.5/3MiniMax M3.1 Flash (free route): 2.5 of 3 checks | 2.5/3Qwen3.8 Max (0902): 2.5 of 3 checks | 1.5/3Gemini 3.8 Flash: 1.5 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| The Starry Night, redrawn in SVG | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4Claude Fable 5.1: 4 of 4 checks | 4/4Kimi K3: 4 of 4 checks | 4/4★GPT-6.1 Sol: 4 of 4 checks, our pick | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 2.5/4DeepSeek V4.1 Flash 1017: 2.5 of 4 checks | 2.5/4Muse Spark 1.3: 2.5 of 4 checks | 1.5/4DeepSeek V4.1 Flash: 1.5 of 4 checks | 3/4GPT-6 Luna: 3 of 4 checks | 2.5/4Claude Haiku 5.5: 2.5 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 3/4Mercury 3: 3 of 4 checks | 2.5/4Mistral Large 4: 2.5 of 4 checks | 1.5/4Grok 4.7: 1.5 of 4 checks | 3/4MiniMax M3.1 Flash (free route): 3 of 4 checks | 2.5/4Qwen3.8 Max (0902): 2.5 of 4 checks | 3/4Gemini 3.8 Flash: 3 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Ball bouncing in a spinning hexagon | 3/3Claude Fable 5.2: 3 of 3 checks | 3/3Claude Fable 5.1: 3 of 3 checks | 3/3Kimi K3: 3 of 3 checks | 3/3GPT-6.1 Sol: 3 of 3 checks | 3/3Kimi K3.1 Pro: 3 of 3 checks | 3/3★Claude Opus 5.5: 3 of 3 checks, our pick | 1.5/3DeepSeek V4.1 Flash 1017: 1.5 of 3 checks | 3/3Muse Spark 1.3: 3 of 3 checks | 3/3DeepSeek V4.1 Flash: 3 of 3 checks | 3/3GPT-6 Luna: 3 of 3 checks | 3/3Claude Haiku 5.5: 3 of 3 checks | –Muse Wonder Full: did not take this test | 3/3GPT-6.1 Terra: 3 of 3 checks | 3/3Mercury 3: 3 of 3 checks | 3/3Mistral Large 4: 3 of 3 checks | 3/3Grok 4.7: 3 of 3 checks | 3/3MiniMax M3.1 Flash (free route): 3 of 3 checks | 3/3Qwen3.8 Max (0902): 3 of 3 checks | 2.5/3Gemini 3.8 Flash: 2.5 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| The solar system, to scale in time | 3/3Claude Fable 5.2: 3 of 3 checks | 3/3Claude Fable 5.1: 3 of 3 checks | 3/3Kimi K3: 3 of 3 checks | 3/3GPT-6.1 Sol: 3 of 3 checks | 3/3Kimi K3.1 Pro: 3 of 3 checks | 3/3★Claude Opus 5.5: 3 of 3 checks, our pick | 3/3DeepSeek V4.1 Flash 1017: 3 of 3 checks | 3/3Muse Spark 1.3: 3 of 3 checks | 3/3DeepSeek V4.1 Flash: 3 of 3 checks | 3/3GPT-6 Luna: 3 of 3 checks | 3/3Claude Haiku 5.5: 3 of 3 checks | –Muse Wonder Full: did not take this test | 3/3GPT-6.1 Terra: 3 of 3 checks | 3/3Mercury 3: 3 of 3 checks | 3/3Mistral Large 4: 3 of 3 checks | 3/3Grok 4.7: 3 of 3 checks | 3/3MiniMax M3.1 Flash (free route): 3 of 3 checks | 3/3Qwen3.8 Max (0902): 3 of 3 checks | 3/3Gemini 3.8 Flash: 3 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| A product page from one feed row | 5/5Claude Fable 5.2: 5 of 5 checks | 5/5Claude Fable 5.1: 5 of 5 checks | 5/5Kimi K3: 5 of 5 checks | 5/5GPT-6.1 Sol: 5 of 5 checks | 5/5Kimi K3.1 Pro: 5 of 5 checks | 5/5★Claude Opus 5.5: 5 of 5 checks, our pick | 5/5DeepSeek V4.1 Flash 1017: 5 of 5 checks | 5/5Muse Spark 1.3: 5 of 5 checks | 5/5DeepSeek V4.1 Flash: 5 of 5 checks | 5/5GPT-6 Luna: 5 of 5 checks | 5/5Claude Haiku 5.5: 5 of 5 checks | –Muse Wonder Full: did not take this test | 5/5GPT-6.1 Terra: 5 of 5 checks | 5/5Mercury 3: 5 of 5 checks | 4/5Mistral Large 4: 4 of 5 checks | 4.5/5Grok 4.7: 4.5 of 5 checks | 5/5MiniMax M3.1 Flash (free route): 5 of 5 checks | 5/5Qwen3.8 Max (0902): 5 of 5 checks | 5/5Gemini 3.8 Flash: 5 of 5 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| A pricing table with a monthly/yearly switch | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4Claude Fable 5.1: 4 of 4 checks | 4/4★Kimi K3: 4 of 4 checks, our pick | 4/4GPT-6.1 Sol: 4 of 4 checks | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | 4/4Muse Spark 1.3: 4 of 4 checks | 4/4DeepSeek V4.1 Flash: 4 of 4 checks | 3.5/4GPT-6 Luna: 3.5 of 4 checks | 4/4Claude Haiku 5.5: 4 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 4/4Mercury 3: 4 of 4 checks | 3.5/4Mistral Large 4: 3.5 of 4 checks | 4/4Grok 4.7: 4 of 4 checks | 4/4MiniMax M3.1 Flash (free route): 4 of 4 checks | 3/4Qwen3.8 Max (0902): 3 of 4 checks | 4/4Gemini 3.8 Flash: 4 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Pelican riding a bicycle | 3/3Claude Fable 5.2: 3 of 3 checks | 3/3Claude Fable 5.1: 3 of 3 checks | 3/3Kimi K3: 3 of 3 checks | 3/3★GPT-6.1 Sol: 3 of 3 checks, our pick | 3/3Kimi K3.1 Pro: 3 of 3 checks | 3/3Claude Opus 5.5: 3 of 3 checks | 3/3DeepSeek V4.1 Flash 1017: 3 of 3 checks | 3/3Muse Spark 1.3: 3 of 3 checks | 2/3DeepSeek V4.1 Flash: 2 of 3 checks | 2.5/3GPT-6 Luna: 2.5 of 3 checks | 2/3Claude Haiku 5.5: 2 of 3 checks | –Muse Wonder Full: did not take this test | 2.5/3GPT-6.1 Terra: 2.5 of 3 checks | 3/3Mercury 3: 3 of 3 checks | 3/3Mistral Large 4: 3 of 3 checks | 2/3Grok 4.7: 2 of 3 checks | 2.5/3MiniMax M3.1 Flash (free route): 2.5 of 3 checks | 3/3Qwen3.8 Max (0902): 3 of 3 checks | 2.5/3Gemini 3.8 Flash: 2.5 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Newton's cradle | 3/3Claude Fable 5.2: 3 of 3 checks | 3/3Claude Fable 5.1: 3 of 3 checks | 3/3Kimi K3: 3 of 3 checks | 2.5/3GPT-6.1 Sol: 2.5 of 3 checks | 3/3Kimi K3.1 Pro: 3 of 3 checks | 3/3★Claude Opus 5.5: 3 of 3 checks, our pick | 3/3DeepSeek V4.1 Flash 1017: 3 of 3 checks | 3/3Muse Spark 1.3: 3 of 3 checks | 3/3DeepSeek V4.1 Flash: 3 of 3 checks | 3/3GPT-6 Luna: 3 of 3 checks | 1/3Claude Haiku 5.5: 1 of 3 checks | –Muse Wonder Full: did not take this test | 3/3GPT-6.1 Terra: 3 of 3 checks | 3/3Mercury 3: 3 of 3 checks | 2.5/3Mistral Large 4: 2.5 of 3 checks | 3/3Grok 4.7: 3 of 3 checks | 3/3MiniMax M3.1 Flash (free route): 3 of 3 checks | 3/3Qwen3.8 Max (0902): 3 of 3 checks | 3/3Gemini 3.8 Flash: 3 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| A newsletter email that works in email | 4.5/5Claude Fable 5.2: 4.5 of 5 checks | 4/5Claude Fable 5.1: 4 of 5 checks | 4/5Kimi K3: 4 of 5 checks | 4.5/5★GPT-6.1 Sol: 4.5 of 5 checks, our pick | 4.5/5Kimi K3.1 Pro: 4.5 of 5 checks | 4.5/5Claude Opus 5.5: 4.5 of 5 checks | 4/5DeepSeek V4.1 Flash 1017: 4 of 5 checks | 3.5/5Muse Spark 1.3: 3.5 of 5 checks | 4/5DeepSeek V4.1 Flash: 4 of 5 checks | 4.5/5GPT-6 Luna: 4.5 of 5 checks | 4/5Claude Haiku 5.5: 4 of 5 checks | –Muse Wonder Full: did not take this test | 4.5/5GPT-6.1 Terra: 4.5 of 5 checks | 4/5Mercury 3: 4 of 5 checks | 4/5Mistral Large 4: 4 of 5 checks | 4/5Grok 4.7: 4 of 5 checks | 4/5MiniMax M3.1 Flash (free route): 4 of 5 checks | 3.5/5Qwen3.8 Max (0902): 3.5 of 5 checks | 3.5/5Gemini 3.8 Flash: 3.5 of 5 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| A landing page for a photo-book app | 3/3Claude Fable 5.2: 3 of 3 checks | 2/3Claude Fable 5.1: 2 of 3 checks | 3/3Kimi K3: 3 of 3 checks | 2.5/3GPT-6.1 Sol: 2.5 of 3 checks | 3/3Kimi K3.1 Pro: 3 of 3 checks | 2.5/3Claude Opus 5.5: 2.5 of 3 checks | 2.5/3DeepSeek V4.1 Flash 1017: 2.5 of 3 checks | 2/3Muse Spark 1.3: 2 of 3 checks | 2/3DeepSeek V4.1 Flash: 2 of 3 checks | 2.5/3GPT-6 Luna: 2.5 of 3 checks | 2.5/3Claude Haiku 5.5: 2.5 of 3 checks | –Muse Wonder Full: did not take this test | 3/3GPT-6.1 Terra: 3 of 3 checks | 2.5/3Mercury 3: 2.5 of 3 checks | 3/3Mistral Large 4: 3 of 3 checks | 3/3★Grok 4.7: 3 of 3 checks, our pick | 3/3MiniMax M3.1 Flash (free route): 3 of 3 checks | 2/3Qwen3.8 Max (0902): 2 of 3 checks | 3/3Gemini 3.8 Flash: 3 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| The jelly effect | 3/3Claude Fable 5.2: 3 of 3 checks | 3/3Claude Fable 5.1: 3 of 3 checks | 2.5/3Kimi K3: 2.5 of 3 checks | 3/3GPT-6.1 Sol: 3 of 3 checks | 2.5/3Kimi K3.1 Pro: 2.5 of 3 checks | 3/3★Claude Opus 5.5: 3 of 3 checks, our pick | 2.5/3DeepSeek V4.1 Flash 1017: 2.5 of 3 checks | 2/3Muse Spark 1.3: 2 of 3 checks | 2.5/3DeepSeek V4.1 Flash: 2.5 of 3 checks | 2.5/3GPT-6 Luna: 2.5 of 3 checks | 1/3Claude Haiku 5.5: 1 of 3 checks | –Muse Wonder Full: did not take this test | 3/3GPT-6.1 Terra: 3 of 3 checks | 2.5/3Mercury 3: 2.5 of 3 checks | 3/3Mistral Large 4: 3 of 3 checks | 3/3Grok 4.7: 3 of 3 checks | 1.5/3MiniMax M3.1 Flash (free route): 1.5 of 3 checks | 3/3Qwen3.8 Max (0902): 3 of 3 checks | 1/3Gemini 3.8 Flash: 1 of 3 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Conway's Game of Life | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4Claude Fable 5.1: 4 of 4 checks | 4/4Kimi K3: 4 of 4 checks | 4/4GPT-6.1 Sol: 4 of 4 checks | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | 3.5/4Muse Spark 1.3: 3.5 of 4 checks | 4/4DeepSeek V4.1 Flash: 4 of 4 checks | 3.5/4GPT-6 Luna: 3.5 of 4 checks | 4/4Claude Haiku 5.5: 4 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 3.5/4Mercury 3: 3.5 of 4 checks | 3/4Mistral Large 4: 3 of 4 checks | 4/4Grok 4.7: 4 of 4 checks | 3.5/4MiniMax M3.1 Flash (free route): 3.5 of 4 checks | 3.5/4Qwen3.8 Max (0902): 3.5 of 4 checks | 4/4Gemini 3.8 Flash: 4 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Flappy Bird in one file | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4Claude Fable 5.1: 4 of 4 checks | 4/4Kimi K3: 4 of 4 checks | 4/4★GPT-6.1 Sol: 4 of 4 checks, our pick | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | 4/4Muse Spark 1.3: 4 of 4 checks | 4/4DeepSeek V4.1 Flash: 4 of 4 checks | 4/4GPT-6 Luna: 4 of 4 checks | 4/4Claude Haiku 5.5: 4 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 4/4Mercury 3: 4 of 4 checks | 4/4Mistral Large 4: 4 of 4 checks | 4/4Grok 4.7: 4 of 4 checks | 4/4MiniMax M3.1 Flash (free route): 4 of 4 checks | 4/4Qwen3.8 Max (0902): 4 of 4 checks | 4/4Gemini 3.8 Flash: 4 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| Chess with every rule | 5/5Claude Fable 5.2: 5 of 5 checks | 5/5Claude Fable 5.1: 5 of 5 checks | 5/5Kimi K3: 5 of 5 checks | 5/5★GPT-6.1 Sol: 5 of 5 checks, our pick | 5/5Kimi K3.1 Pro: 5 of 5 checks | 5/5Claude Opus 5.5: 5 of 5 checks | 5/5DeepSeek V4.1 Flash 1017: 5 of 5 checks | 5/5Muse Spark 1.3: 5 of 5 checks | 5/5DeepSeek V4.1 Flash: 5 of 5 checks | 5/5GPT-6 Luna: 5 of 5 checks | 5/5Claude Haiku 5.5: 5 of 5 checks | –Muse Wonder Full: did not take this test | 5/5GPT-6.1 Terra: 5 of 5 checks | 5/5Mercury 3: 5 of 5 checks | 5/5Mistral Large 4: 5 of 5 checks | 5/5Grok 4.7: 5 of 5 checks | 5/5MiniMax M3.1 Flash (free route): 5 of 5 checks | 5/5Qwen3.8 Max (0902): 5 of 5 checks | 0/5Gemini 3.8 Flash: 0 of 5 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| A category page from a product feed | 4/4Claude Fable 5.2: 4 of 4 checks | 3.5/4Claude Fable 5.1: 3.5 of 4 checks | 3.5/4Kimi K3: 3.5 of 4 checks | 4/4GPT-6.1 Sol: 4 of 4 checks | 3.5/4Kimi K3.1 Pro: 3.5 of 4 checks | 4/4★Claude Opus 5.5: 4 of 4 checks, our pick | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | 3/4Muse Spark 1.3: 3 of 4 checks | 4/4DeepSeek V4.1 Flash: 4 of 4 checks | 4/4GPT-6 Luna: 4 of 4 checks | 4/4Claude Haiku 5.5: 4 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 4/4Mercury 3: 4 of 4 checks | 3/4Mistral Large 4: 3 of 4 checks | 0/4Grok 4.7: 0 of 4 checks | 4/4MiniMax M3.1 Flash (free route): 4 of 4 checks | 3/4Qwen3.8 Max (0902): 3 of 4 checks | 3/4Gemini 3.8 Flash: 3 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| An animated SVG logo | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4★Claude Fable 5.1: 4 of 4 checks, our pick | 4/4Kimi K3: 4 of 4 checks | 4/4GPT-6.1 Sol: 4 of 4 checks | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | 4/4Muse Spark 1.3: 4 of 4 checks | 4/4DeepSeek V4.1 Flash: 4 of 4 checks | 4/4GPT-6 Luna: 4 of 4 checks | 4/4Claude Haiku 5.5: 4 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 4/4Mercury 3: 4 of 4 checks | 4/4Mistral Large 4: 4 of 4 checks | 4/4Grok 4.7: 4 of 4 checks | 4/4MiniMax M3.1 Flash (free route): 4 of 4 checks | 4/4Qwen3.8 Max (0902): 4 of 4 checks | 3.5/4Gemini 3.8 Flash: 3.5 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |
| An analogue clock that tells the time | 4/4Claude Fable 5.2: 4 of 4 checks | 4/4Claude Fable 5.1: 4 of 4 checks | 4/4Kimi K3: 4 of 4 checks | 4/4★GPT-6.1 Sol: 4 of 4 checks, our pick | 4/4Kimi K3.1 Pro: 4 of 4 checks | 4/4Claude Opus 5.5: 4 of 4 checks | 4/4DeepSeek V4.1 Flash 1017: 4 of 4 checks | 4/4Muse Spark 1.3: 4 of 4 checks | 4/4DeepSeek V4.1 Flash: 4 of 4 checks | 2/4GPT-6 Luna: 2 of 4 checks | 4/4Claude Haiku 5.5: 4 of 4 checks | –Muse Wonder Full: did not take this test | 4/4GPT-6.1 Terra: 4 of 4 checks | 4/4Mercury 3: 4 of 4 checks | 3/4Mistral Large 4: 3 of 4 checks | 3.5/4Grok 4.7: 3.5 of 4 checks | 4/4MiniMax M3.1 Flash (free route): 4 of 4 checks | 3.5/4Qwen3.8 Max (0902): 3.5 of 4 checks | 3.5/4Gemini 3.8 Flash: 3.5 of 4 checks | –Muse Wonder: did not take this test | ·Gemini 4 Argon: not checked |