The answers to “An analogue clock that tells the time”, side by side

Playground · Drawing · HTML

An analogue clock that tells the time

AI models each build a beautiful analogue clock showing local time. We check the time, the second hand, the hour hand and the dial.

About this test

An analogue clock looks simple, but it is easy to get quietly wrong: the hour hand has to creep between the hours as the minutes pass, the second hand has to move, and every hour has to be on the dial. We checked each clock against the real time, a second apart, and again at a fixed 10:42:30 to see where the hour hand sits.

Read more

The prompt, word for word

Create an analogue clock that shows the current local time, with hour, minute and second hands and the hours marked around the dial. Make it look beautiful.

Make it a single self-contained HTML file (inline CSS and JavaScript, no external libraries, fonts or images).
Show the whole prompt

Default settings: no thinking effort set, no temperature, and room for 64,000 tokens. A card says so when an answer was made with other settings: “limit 128k” means it could use up to 128,000.

Every answer runs live, in a sandboxed frame. Click into one to play with it.

How each model did

RankModelChecksTimeCost
1 GPT-6.1 Sol Our pick 4 of 4 54 s $0.034
2 Claude Haiku 5.5 Best value 4 of 4 30 s $0.003
3 Mercury 3 4 of 4 36 s $0.005
4 MiniMax M3.1 Flash (free route) 4 of 4 114 s $0.010
5 DeepSeek V4.1 Flash 4 of 4 28 s $0.010
6 GPT-6.1 Terra 4 of 4 31 s $0.014
7 Muse Spark 1.3 4 of 4 76 s $0.025
8 DeepSeek V4.1 Flash 1017 4 of 4 80 s $0.026
9 Claude Opus 5.5 4 of 4 62 s $0.145
10 Kimi K3.1 Pro 4 of 4 7.8 min $0.175
11 Kimi K3 4 of 4 3.4 min $0.178
12 Claude Fable 5.1 4 of 4 41 s $0.188
13 Claude Fable 5.2 4 of 4 115 s $0.454
14 Grok 4.7 3.5 of 4 23 s $0.013
15 Gemini 3.8 Flash 3.5 of 4 55 s $0.025
16 Gemini 4 Carbon 3.5 of 4 83 s $0.067
17 Qwen3.8 Max (0902) 3.5 of 4 12 min $0.172
18 Mistral Large 4 3 of 4 2.6 min $0.013
19 GPT-6 Luna 2 of 4 25 s $0.001
20 Muse Ember Pending 17 s $0.005
21 Gemini 4 Argon Pending 29 s $0.045

Ranked by our pick, then checks passed, then cost. Time and cost count every try; each dot is one check, in the order listed below (filled: passed, half: partly, empty: failed). Click a model for its notes.

What each answer cost, against the checks it passed

Best value: Claude Haiku 5.5, 4 of 4 checks for $0.003. Our pick: GPT-6.1 Sol, 4 of 4 for $0.034. The best value is the cheapest answer with the most checks passed, counting every try.

0%20%40%60%80%100% $0.001$0.01$0.1$1 Cost, every try (log scale) Checks passed GPT-6.1 Sol Claude Haiku 5.5

Best valueOur pickOther answers

Notes on each answer

1

GPT-6.1 Sol

Our pick 4 of 4

A cream dial in a brass bezel on a deep green page headed "The present hour.", with serif numerals, a digital readout, the date and the time zone underneath. The hands were right to the second against the real clock, and at 10:42 the hour hand sits about two thirds of the way from 10 to 11. Nothing missing and nothing out of place.

54 s · 3,422 output tokens (302 thinking) · $0.034

2

Claude Haiku 5.5

4 of 4

A navy dial with cream Roman numerals, gold baton markers, a gold bezel, sword-shaped hands and a red second hand, with a digital readout and the date underneath. The time matched the real clock to the second, and at 10:42 the hour hand sits about two thirds of the way from X to XI. The hour markers touch the edges of VIII and IV, but every numeral is readable. The cheapest of the first eleven answers, at $0.0034.

30 s · 6,799 output tokens (1,814 thinking) · $0.003

Added 8 October 2026.
3

Mercury 3

4 of 4

A cream dial in a silver bezel on a dark page, with navy numerals and baton markers, minute ticks, navy hands and a thin red second hand that sweeps. It matched the real clock to the second, and at 1:31 the hour hand sat halfway between 1 and 2. There is no digital readout or date. 36 seconds, $0.0048.

36 s · 13,581 output tokens (11,626 thinking) · $0.005

Added 19 October 2026.
4

MiniMax M3.1 Flash (free route)

4 of 4

A navy dial in a gold bezel on a dark page, with serif numerals, gold hour batons, a gold hour hand, a silver minute hand and a thin red second hand that sweeps smoothly, plus a large digital readout, the date and the time zone underneath. It matched the real clock to the second, and at 10:03 the hour hand sat just past the 10. The small labels on the dial ("Local / 24 hour", "Every second") make little sense on a twelve-hour face. One try, 114 seconds on OpenRouter's free route, so the time is not representative; $0.0097 at list prices.

114 s · 16,106 output tokens (11,408 thinking) · $0.010

Added 19 October 2026.
5

DeepSeek V4.1 Flash

4 of 4

A simple dark dial with a gold rim, white numerals and hands, and an orange second hand. Every check passes and the dial is easy to read, though it is plainer than the best answers. The cheapest of the first ten answers, at under a cent.

28 s · 8,213 output tokens (5,894 thinking) · $0.010

6

GPT-6.1 Terra

4 of 4

A dark green dial in a slim brass ring, under the heading "Here & Now" with the date below it. It told the right time, the second hand moves, and with the clock set to 4:30 the hour hand sat halfway between 4 and 5. All twelve hours are clear serif numerals. The heading and date take up room, so the dial is smaller than it could be. 31 seconds, $0.014.

31 s · 2,734 output tokens (319 thinking) · $0.014

Added 27 October 2026.
7

Muse Spark 1.3

4 of 4

A navy and gold "AURUM" dial with bold white numerals, a pink second hand, a date pill and a digital readout with the time zone. The time was correct and the hour hand moves in proportion. The date pill sits under the hands near the centre, so the hands cross it.

76 s · 5,899 output tokens (2,391 thinking) · $0.025

8

DeepSeek V4.1 Flash 1017

4 of 4

A navy dial in a brushed gold bezel on a dark page, with cream serif numerals, gold baton markers, tapered gold hands and a thin red second hand that sweeps. It matched the real clock to the second, and the hour hand sits in proportion (at 4:34 it was just past halfway from 4 to 5). There is no digital readout or date. 80 seconds, $0.026.

80 s · 21,505 output tokens (18,608 thinking) · $0.026

Added 19 October 2026.
9

Claude Opus 5.5

4 of 4

A navy dial with gold Roman numerals (a watchmaker's IIII for four), a gold rim, a red second hand and a date window, labelled "TEMPUS". The time was exact and the hour hand moves between hours in proportion. It is as polished as GPT-6.1's clock, at about four times the cost.

62 s · 7,237 output tokens (847 thinking) · $0.145

10

Kimi K3.1 Pro

4 of 4

A navy dial in a heavy gold case, with a red second hand and a small "Local time" label. Right time, a moving second hand, and at 4:30 the hour hand sat halfway between 4 and 5. It fills the frame, though the minute hand is long enough to run over the numerals, and it has no date. 7.8 minutes, $0.18.

7.8 min · 17,518 output tokens (13,591 thinking) · $0.175

Added 27 October 2026.
11

Kimi K3

4 of 4

A navy "MERIDIAN" dial with gold Roman numerals (IIII for four), a heavy polished gold bezel and a red second hand. The time was correct and the hour hand sits in proportion between hours. It looks very close to Claude Opus 5.5's clock and took over three minutes.

3.4 min · 13,695 output tokens (9,596 thinking) · $0.178

12

Claude Fable 5.1

4 of 4

A classic cream dial with black Roman numerals, a gold case and a red second hand with a counterweight, labelled "CHRONOS". The hands showed the correct time and the hour hand sat halfway between IIII and V at 4:28. Clean and correct, but the most expensive of the first eighteen answers on the page.

41 s · 3,750 output tokens (142 thinking) · $0.188

13

Claude Fable 5.2

4 of 4

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.

115 s · 11,334 output tokens (6,069 thinking) · $0.454

Added 10 November 2026.
14

Grok 4.7

3.5 of 4

A plain white dial with black stick hands, a thin red second hand and a dense ring of minute ticks. The time is right, but the long hour ticks run straight through the numerals, so 9, 3, 4 and 8 are struck through. One of the fastest answers, at 23 seconds.

  • All twelve hours are marked and readable (partly)

23 s · 2,084 output tokens (54 thinking) · $0.013

Billed for 1,297 input tokens, about 1,210 more than the other models for the same prompt. We sent no system prompt, and the extra is there on every one of its requests, so it is text the provider adds itself.
15

Gemini 3.8 Flash

3.5 of 4

A dark "CHRONOMETER" dial with thin gold hands, small gold numerals and a date window. The time and hands are right, but the date window replaces the 3, so only eleven hours are numbered, and the numerals are small for the size of the dial.

  • All twelve hours are marked and readable (partly)

55 s · 6,772 output tokens (2,447 thinking) · $0.025

16

Gemini 4 Carbon

3.5 of 4

A navy watch dial in a silver bezel, with serif numerals, a date window and a red sweeping second hand. At 10:42:30 the hand angles were right and the second hand advanced six degrees in a second. The 11 and 12 overlap the CHRONOMÈTRE label, so the dial is less readable than it first looks. 1 try, 83 s, $0.0666.

  • All twelve hours are marked and readable (partly)

83 s · 6,638 output tokens (1,400 thinking) · $0.067

Added 11 November 2026.
17

Qwen3.8 Max (0902)

3.5 of 4

An ornate cream dial with blue steel hands, a date window, a day-progress bar and a sweep/tick switch. The hands are right, but the large Roman numerals are crowded on a small ring, with XI, XII and I running into the "MERIDIAN" label, and the date window takes the place of III. It took over 12 minutes.

  • All twelve hours are marked and readable (partly)

12 min · 28,697 output tokens (20,805 thinking) · $0.172

18

Mistral Large 4

3 of 4

A glowing dark blue dial with gold, white and pink hands. The hands show the right time, but the dial is numbered in minutes (12, 5, 10 ... 55) instead of hours, so 4:29 reads as the hour hand at "20" and the minute hand at "30".

  • All twelve hours are marked and readable

2.6 min · 5,999 output tokens (2,958 thinking) · $0.013

19

GPT-6 Luna

2 of 4

A cream dial with serif numerals in a brass bezel on a dark card headed "Local time", with a digital readout and the date underneath. The readout was right to the second, but the hands are not: each one turns about the bottom-right corner of the drawing instead of the centre, so at 10:42 the hour hand lay near 7 and the minute and second hands were off the dial, and the second hand spends most of each minute sweeping across the page outside the clock. The hands only line up near 12 o'clock. The cheapest of the first twelve answers, at $0.0014.

  • Shows the correct current time (within a minute) (partly)
  • The second hand moves (ticking or sweeping) (partly)
  • The hour hand sits between hours in proportion to the minutes (at 4:30 it is halfway from 4 to 5)

25 s · 2,880 output tokens (740 thinking) · $0.001

Added 8 October 2026.
20

Muse Ember

17 s · 9,257 output tokens (6,338 thinking) · $0.005

Added 9 November 2026.
21

Gemini 4 Argon

29 s · 4,441 output tokens · $0.045

Added 13 October 2026.

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. This is a quick look, not a benchmark: run it again and every answer will differ. For measured results, see our benchmarks.

How we pick. Claude Opus 5.5 tried every answer (clicking, dragging and playing it in a browser), marked the checks and drafted the notes. Our pick is a judgement among the answers with the most checks passed, not a score: the checks show what works, the pick is which of those we would rather use. Opus is an Anthropic model, and Claude answers are our pick in 18 of 29 tests, so weigh Claude picks with that in mind. Our method calls for a blind second opinion from another lab's model on any Claude pick; none is recorded for this page yet.

What we checked, by trying each answer ourselves: Shows the correct current time (within a minute); The second hand moves (ticking or sweeping); The hour hand sits between hours in proportion to the minutes (at 4:30 it is halfway from 4 to 5); All twelve hours are marked and readable. A partial pass counts as half.

Verdict

All but one clock told the right time, had a moving second hand and moved the hour hand between hours properly (GPT-6 Luna's hands turn about the wrong point), so the dial and the finish made the difference. GPT-6.1 Sol's brass and cream clock with its date and time-zone readout is the best overall, and at 54 seconds and $0.034 it costs a quarter of what Claude Opus 5.5 ($0.145), Claude Fable 5.1 ($0.188) and Kimi K3 ($0.178) charged for equally correct clocks. DeepSeek V4.1 Flash gave a plain but fully correct clock in 28 seconds for $0.0099. DeepSeek V4.1 Flash 1017, added on 19 October, is also fully correct, with a darker gold-rimmed dial, in 80 seconds for $0.026. MiniMax M3.1 Flash, added the same day on OpenRouter's free route, is correct as well, with a navy and gold dial and a digital readout underneath, and so is Mercury 3, added the same day, with a plain cream dial, in 36 seconds. GPT-6.1 Terra, added on 27 October, is fully correct too, with a dark green dial, in 31 seconds for $0.014. Kimi K3.1 Pro, added the same day, is also correct, with a navy dial in a heavy gold case, but it took 7.8 minutes and cost $0.18. Claude Fable 5.2, added on 10 November, is fully correct as well, with a cream dial in a steel bezel and a date window, in 115 seconds, but at $0.45 it is the dearest of the first nineteen answers.

The faults were all on the dial. Gemini 3.8 Flash and Qwen3.8 Max put a date window where the 3 should be, Grok 4.7 ran its hour ticks through the numerals, and Mistral Large 4 numbered the dial in minutes, so it shows no hours at all. Qwen took 12 minutes.

What would your prompt get?The Spring Prompt playground runs one prompt on several models at once, from OpenAI, Anthropic, Google, Mistral and others, side by side.
Ask for access
More from the playground
All tests →
Export
Include

Chrome will ask to share this tab: choose Allow. Only the board below is recorded.

Spring Promptspringprompt.com