The answers to “Flappy Bird in one file”, side by side

Playground · Game · HTML

Flappy Bird in one file

AI models each build Flappy Bird in one HTML file. We played every game: flapping, pipes, scoring, crashing and restarting.

About this test

Flappy Bird is a small game with little room for error: gravity and the flap have to feel right, the pipes need gaps you can actually get through, the score has to tick up once per pipe, and a crash has to end the run and let you go again straight away. We played every answer with a script that flaps with Space and clicks, steers through the gaps, lets the bird fall, and restarts. Look at how each game handles the game-over screen and the restart, which is where the answers differ most.

Read more

The prompt, word for word

Make a Flappy Bird game: tap, click or press space to flap, pipes scroll past with gaps to fly through, the score goes up for each pipe passed, and hitting a pipe or the ground ends the game with a way to restart.

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 84 s $0.054
2 Mercury 3 Best value 4 of 4 8 s $0.002
3 GPT-6 Luna 4 of 4 47 s $0.003
4 Claude Haiku 5.5 4 of 4 55 s $0.006
5 DeepSeek V4.1 Flash 4 of 4 2.2 min $0.015
6 MiniMax M3.1 Flash (free route) 4 of 4 4.3 min $0.018
7 Grok 4.7 4 of 4 33 s $0.019
8 GPT-6.1 Terra 4 of 4 44 s $0.020
9 Gemini 3.8 Flash 4 of 4 52 s $0.027
10 Muse Spark 1.3 4 of 4 3.4 min $0.033
11 DeepSeek V4.1 Flash 1017 4 of 4 103 s $0.036
12 Gemini 4 Carbon 4 of 4 106 s $0.088
13 Mistral Large 4 4 of 4 11 min $0.126
14 Claude Opus 5.5 4 of 4 72 s $0.176
15 Kimi K3 4 of 4 89 s $0.198
16 Kimi K3.1 Pro 4 of 4 10 min $0.218
17 Qwen3.8 Max (0902) 4 of 4 23 min $0.325
18 Claude Fable 5.1 4 of 4 81 s $0.392
19 Claude Fable 5.2 4 of 4 3.3 min $0.856
20 Muse Ember Pending 13 s $0.004
21 Gemini 4 Argon Pending 34 s $0.054

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: Mercury 3, 4 of 4 checks for $0.002. Our pick: GPT-6.1 Sol, 4 of 4 for $0.054. 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 Mercury 3

Best valueOur pickOther answers

Notes on each answer

1

GPT-6.1 Sol

Our pick 4 of 4

A restyled take called "little flap.", with soft green pipes, rolling hills, a sun and a "Best" badge in the corner. It plays well: we passed 11 pipes, and the crash brings up a "Nice flight!" card with score and best and a "Fly again" button that resets cleanly. The only answer with its own look rather than a copy of the original.

84 s · 5,356 output tokens (516 thinking) · $0.054

2

Mercury 3

4 of 4

A faithful classic: a yellow bird, green pipes, drifting clouds and a striped ground, with a "Flappy" start card. Clicks, taps and Space all flap and gravity feels right; our steering script passed pipes and the score went up one for each, and hitting a pipe or the ground brings up "Game Over" with the score. A tap restarts at 0 straight away, with no pause and no best score. The fastest answer on the page, at 8 seconds, $0.0016.

8 s · 4,630 output tokens (926 thinking) · $0.002

Added 19 October 2026.
3

GPT-6 Luna

4 of 4

A bright classic with an orange bird, soft green pipes, a sun, hills and a sandy ground, a "FLAPPY BIRD" title card and a best score. Space, clicks and taps all flap and gravity feels right; our steering script passed 36 pipes, scoring one for each, and hitting a pipe or the ground ends the run under "GAME OVER" with score and best. "Play again", Space or a tap starts a new run at 0 with a flap, but there is no ready pause, so a tap just after a crash restarts at once. The cheapest of the first twelve answers, at $0.0028.

47 s · 5,632 output tokens (1,339 thinking) · $0.003

Added 8 October 2026.
4

Claude Haiku 5.5

4 of 4

A faithful classic with a yellow bird, green pipes, drifting clouds and a striped ground under a "FLAPPY BIRD" title. Flapping and gravity feel right; we passed 12 pipes, and a crash brings up a Game Over card with score and best. A short pause stops an accidental restart, then a tap or Space starts a new run at 0 with a flap. The clouds have odd sharp spikes on their edges. The cheapest of the first eleven answers, at $0.0065.

55 s · 12,919 output tokens (6,027 thinking) · $0.006

Added 8 October 2026.
5

DeepSeek V4.1 Flash

4 of 4

A tidy classic with a bright sky, tall pipes and a white rounded "GAME OVER" card with score and best. We passed 10 pipes, and the restart goes to a "TAP TO START" screen. The cheapest of the first ten answers, and it works fully.

2.2 min · 26,759 output tokens (19,270 thinking) · $0.015

6

MiniMax M3.1 Flash (free route)

4 of 4

"Wing It": a golden bird in a rounded portrait frame, with green pipes, clouds, hills and a sun, a best score and a start card. Clicks, taps and Space all flap, and the gaps narrow and the pipes speed up as the score rises; our steering script passed 59 pipes, and a crash brings up a card with the pipes cleared, the best and a "New personal best" badge. A tap or Space starts a new run at 0 with a flap. Touching the top of the screen also ends the run, which the original does not do, and there is no sound. One try, 4.3 minutes on OpenRouter's free route, so the time is not representative; $0.018 at list prices.

4.3 min · 29,958 output tokens (20,686 thinking) · $0.018

Added 19 October 2026.
7

Grok 4.7

4 of 4

A plain version with flat clouds and a hatched ground; a crash dims the screen under "Game over", the score and a yellow Restart button. Flapping, pipes and scoring work (8 pipes on our best run). Space restarts with the bird already falling and no ready pause, so it can hit the ground before you react, and tapping the game after a crash does nothing.

33 s · 3,091 output tokens (139 thinking) · $0.019

Billed for 1,316 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.
8

GPT-6.1 Terra

4 of 4

A faithful copy in a white-edged panel: yellow bird, green pipes and a "Ready to fly?" card. Space flaps, gravity pulls the bird down and each pipe scores one: we passed nine. A crash brings up "Game over!" with score and best, and "Play again" starts at 0. Touching the top of the screen ends the run. 44 seconds, $0.020.

44 s · 3,907 output tokens (774 thinking) · $0.020

Added 27 October 2026.
9

Gemini 3.8 Flash

4 of 4

A city-skyline Flappy Bird with chunky "GAME OVER" lettering, a medal slot and little sparkles when the bird flaps. Everything works, but the gaps are tighter than in the other games, so our best was 3 pipes against 6 to 11 elsewhere. Tapping after a crash starts a new run straight away at 0.

52 s · 7,304 output tokens (1,188 thinking) · $0.027

10

Muse Spark 1.3

4 of 4

A bright version with an orange bird, puffy clouds and a cream Game Over card with a "NEW BEST!" ribbon and an orange Restart button. We passed 11 pipes; the screen flashes white on the crash, and restarting begins again at 0. One of the cleanest all-round games.

3.4 min · 7,757 output tokens (2,609 thinking) · $0.033

11

DeepSeek V4.1 Flash 1017

4 of 4

A classic in a portrait frame: a yellow bird, green pipes, clouds, hills, a striped ground and a big outlined score, with sounds for the flap, each point and the crash. Clicks, taps and Space all flap; our steering script passed 12 pipes, and hitting a pipe or the ground brings up a "Game Over" card with score and best. After a short pause, a tap starts a new run at 0. 103 seconds, $0.036.

103 s · 30,363 output tokens (24,625 thinking) · $0.036

Added 19 October 2026.
12

Gemini 4 Carbon

4 of 4

A compact cyan game with a round yellow bird and green pipes. Space flaps against gravity; playing through two pipes raised the displayed score to two. A collision ended the game and space restarted it with the score back at zero. The drawing and controls stayed clear. 1 try, 106 s, $0.0884.

106 s · 8,816 output tokens (1,271 thinking) · $0.088

Added 11 November 2026.
13

Mistral Large 4

4 of 4

A pixel-font version with an orange bird, green hills, a mute button and a dark Game Over box with a "PLAY AGAIN" button. It works, though our best run was 6 pipes, fewer than in most of the games; Space or the button restarts at 0.

11 min · 60,131 output tokens (41,727 thinking) · $0.126

14

Claude Opus 5.5

4 of 4

A faithful classic Flappy Bird with green pipes, clouds, a striped ground and a "Sound on (M)" toggle. Flapping and gravity feel right; we reached 11 pipes, and the Game Over panel shows score, best and a "NEW!" ribbon. Space returns to the title screen and the next run starts at 0.

72 s · 8,793 output tokens (932 thinking) · $0.176

15

Kimi K3

4 of 4

A classic look with a "Get Ready!" title screen and a Game Over panel showing score, best and "NEW!". The pipes are spaced wider apart than in most games, so it plays more slowly; we got to 7 before crashing. Restarting goes back to the ready screen.

89 s · 14,648 output tokens (9,474 thinking) · $0.198

16

Kimi K3.1 Pro

4 of 4

Another faithful copy, with a sun and hills behind a rounded frame and an orange bird that tilts as it climbs. Flapping, scoring and crashes all work: we passed nine pipes, and "Game Over" shows score and best with a new-best line before "Fly again" starts at 0. The top of the screen stops the bird without ending the run. 10 minutes, $0.22.

10 min · 21,722 output tokens (15,013 thinking) · $0.218

Added 27 October 2026.
17

Qwen3.8 Max (0902)

4 of 4

The most complete game: a "PIXEL EDITION" with pixel digits, "+1" pop-ups for each pipe, a sky that turns to dusk as the score rises, a medal on the Game Over board, and a P key to pause. We passed 9 pipes, and the restart returns to "GET READY!". It took over 23 minutes to write.

23 min · 54,151 output tokens (40,853 thinking) · $0.325

18

Claude Fable 5.1

4 of 4

Very close to Opus's game: the classic look, a "Flappy Bird" title screen and a Game Over panel with "NEW BEST!". We passed 10 pipes; clicking into the top pipe ended the run at once, and the restart goes back to a ready screen showing "Best: 10". Solid, but the most expensive of the first ten answers.

81 s · 7,823 output tokens (217 thinking) · $0.392

19

Claude Fable 5.2

4 of 4

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.

3.3 min · 21,370 output tokens (9,394 thinking) · $0.856

Added 10 November 2026.
20

Muse Ember

13 s · 7,667 output tokens (3,892 thinking) · $0.004

Added 9 November 2026.
21

Gemini 4 Argon

34 s · 5,385 output tokens · $0.054

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: Clicking, tapping or space makes the bird flap and gravity pulls it down; Pipes scroll past with gaps you can fly through; The score goes up for each pipe passed; Hitting a pipe or the ground ends the game, and it restarts cleanly. A partial pass counts as half.

Verdict

Every model made a working Flappy Bird: each one flaps, falls, scores one point per pipe, ends on a crash and restarts. GPT-6.1 Sol's "little flap." is our pick because it plays well and is the only one with its own design, for $0.054 in 84 seconds. Claude Opus 5.5 and Muse Spark 1.3 are the best of the faithful copies; Muse cost $0.033. Qwen3.8 Max added the most (pause, medals, a dusk sky, "+1" pop-ups), but it took over 23 minutes and $0.33.

For value, DeepSeek V4.1 Flash ($0.015) and Grok 4.7 ($0.019, 33 seconds) both work fully; Grok's restart drops the bird straight away, which feels harsh. DeepSeek V4.1 Flash 1017, added on 19 October, is another faithful copy that works fully, with sound ($0.036). MiniMax M3.1 Flash, added the same day on OpenRouter's free route, works fully too and speeds up as you score, though touching the top ends the run. Gemini 3.8 Flash has tighter gaps and is noticeably harder. Claude Fable 5.1 plays much like Opus but cost the most of the first ten, at $0.39. Mercury 3, added on 19 October, gives another faithful copy that works fully, and was the quickest answer here at 8 seconds. GPT-6.1 Terra, added on 27 October, is a faithful copy that works fully ($0.020), though touching the top ends the run. Kimi K3.1 Pro, added the same day, is another, with a sunny backdrop and a ceiling that only stops the bird ($0.22). Claude Fable 5.2, added on 10 November, is one more faithful copy that works fully, with medals and sound, but it took 3.3 minutes and at $0.86 is the dearest of the first nineteen answers.

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