Playground · Animation · HTML · Standing test
The solar system, to scale in time
An animated solar system is a good small test of whether a model gets the numbers right as well as the look: each planet has to go round at a rate set by its real orbital period, so Mercury laps the Sun about four times per Earth year and Neptune barely moves. We tracked where each planet was drawn over eight seconds and compared how far they went, read the period values in each answer's code, and pressed every speed control.
The prompt, word for word
Make an animated model of the solar system: the Sun and the eight planets on their orbits, each planet orbiting at a speed in proportion to its real orbital period, labelled with its name. Add controls to speed time up and slow it down. Make it a single self-contained HTML file (inline CSS and JavaScript, no external libraries, fonts or images).
Every answer runs live, in a sandboxed frame. Click into one to play with it.
How each model did
| Model | Our notes | Checks | Time | Output tokens | Cost |
|---|---|---|---|---|---|
| GPT-6.1 Sol | A tidy "orbital clock" page with a framed top-down view, all nine bodies labelled, a log-scale speed slider with ½× and 2× buttons, and a row of period cards underneath. Mercury goes round about four times per Earth year, and the speed buttons change the rate as labelled. Plain next to the best answers, but correct and among the cheapest. | 3 of 3 | 67 s | 4,528 313 thinking |
$0.045 |
| Claude Opus 5.5 Our pick | A dark map with all nine bodies labelled, a calendar date at the top right, a legend of real periods, and controls for slower, faster, pause, reverse, orbit spacing and a "Today" button, with planets starting from their real positions for the date. Mercury makes 4.1 laps per Earth year, Reverse runs the calendar backwards, and the spacing button switches to square-root distances. The most complete answer on the page. | 3 of 3 | 91 s | 10,363 1,559 thinking |
$0.208 |
| Claude Fable 5.1 | A clean dark view with real periods and start positions from 1 January 2000, a date readout, slower, faster, reset buttons and a "true distances" switch. Speeds are right (Mercury 4.1 laps per Earth year). In a 1000 by 700 frame the control panel sits over Neptune at the start (it shows in a larger window), and it was the most expensive answer. | 3 of 3 | 93 s | 8,380 3,154 thinking |
$0.420 |
| Gemini 3.8 Flash | A glowing Sun with shaded planets, Saturn with rings, toggles for labels and orbits, zoom buttons and a time-warp slider showing "36.5 days / sec". Speeds are correct (Mercury 4.15 laps per Earth year) and ◄◄/►► halve and double them. The eight planets are labelled but the Sun is not, and they all start lined up in a row. | 3 of 3 | 62 s | 10,680 2,685 thinking |
$0.040 |
| Muse Spark 1.3 | A busy page with a yellow header bar, slower, pause, faster, reset buttons, a speed slider and a strip of real periods along the bottom. Speeds are correct (Mercury about 4.2 laps per Earth year in our tracking) and each button halves or doubles them. A stray line in the footer, "drag? no — top-down view", looks like a note the model left for itself. | 3 of 3 | 159 s | 6,867 2,906 thinking |
$0.029 |
| Grok 4.7 | A simple dark diagram with serif labels for all nine bodies, slower, faster, pause buttons and a speed slider in "Earth-years / min". Speeds are correct (Mercury 4.15 laps per Earth year) and each button press changes speed by about 1.8×. It does the job with the fewest extras, and was the quickest answer at 59 seconds. | 3 of 3 | 59 s | 4,160 755 thinking |
$0.026 |
| Kimi K3 | A large Sun with a wide glow and planets dropped at random starting angles, with slower, pause, faster buttons and a speed slider. The ratios are right (Mercury 4.1 laps per Earth year), but the default 10 days per second is slow, the Sun's glow and label crowd the inner labels, and in a 1000 by 700 frame Neptune can start under the control bar. | 3 of 3 | 71 s | 4,852 463 thinking |
$0.068 |
| Qwen3.8 Max (0902) | The most styled answer: a "mission clock" readout, planets with fading trails, speed presets from a day to a year per second, toggles for orbits, labels and trails, and a dossier you open by clicking a planet. Speeds are correct (Mercury 4.1 laps per Earth year), but the Sun has no label, the Mercury and Venus labels overlap at times, a hover tooltip sits on top of the dossier, and it took almost nine minutes to answer. | 3 of 3 | 525 s | 21,204 10,378 thinking |
$0.127 |
| DeepSeek V4.1 Flash | A minimal, clean view with soft glowing planets and a small bar with reset, minus, pause, plus and a slider showing "1.00×" and elapsed years. Speeds are correct (Mercury 4.15 laps per Earth year) and plus and minus double and halve them. The Sun is not labelled and there is no date or legend, but it is the cheapest answer by far. | 3 of 3 | 99 s | 19,584 14,632 thinking |
$0.011 |
| Mistral Large 4 | A dark view with a control card at the top left: slower, pause, faster, reset positions and a speed slider, where 1× is one Earth year in 12 seconds. Speeds are correct (Mercury 4.15 laps per Earth year) and the buttons work. The inner labels overlap at the start ("Ear" behind "Venus"), in smaller frames the card covers part of the outer orbits including Neptune, and it needed a second try.Took 2 tries: ran out of room at 64,000 tokens; finished at 128,000 tokens. | 3 of 3 | 292 s | 11,802 7,153 thinking |
$0.025 $0.159 all tries |
One answer per model, at each model's default settings unless the card says otherwise, through OpenRouter. If an answer ran out of room or stopped short, we asked again with more room, then with thinking off, and say so. We tried every answer, with Claude Opus 5.5 running the checks and drafting the notes; it is an Anthropic model, so we check its calls on Claude answers ourselves. This is a quick look, not a benchmark: run it again and every answer will differ. For measured results, see our benchmarks.
What we checked, by trying each answer ourselves: The Sun and all eight planets, labelled, in the right order; Orbital speeds are in proportion to the real periods; Speeding up and slowing down works. A partial pass counts as half.
Verdict
Every model got the physics right: all ten use the real periods, and our tracking showed Mercury doing about 4.1 laps per Earth year and Mars about one lap every 1.9 Earth years in each one. So the differences are in presentation and extras. Claude Opus 5.5 is the pick, with real start positions on a calendar date, reverse time, a spacing switch and a period legend, for $0.21 in 91 seconds. Qwen3.8 Max is the richest after it, with trails, presets and click-for-details panels, but it took 525 seconds. GPT-6.1 Sol ($0.045) and DeepSeek V4.1 Flash ($0.011) are correct, clean and cheap.
The weaker answers have layout problems, not wrong numbers. Kimi K3, Claude Fable 5.1 and Mistral Large 4 can hide Neptune behind a control panel in a 1000 by 700 frame, and Fable was the most expensive at $0.42. Mistral's first try ran out of room at 64,000 tokens after 1,125 seconds; the second finished in 292 seconds for $0.025.
What would your prompt get?The Spring Prompt playground runs one prompt on up to six models from OpenAI, Anthropic, Google, Mistral and others, side by side.
Ask for access
More from the playground
