The answers to “Chess with every rule”, side by side

Playground · Game · HTML

Chess with every rule

AI models each build two-player chess in one HTML file. We clicked through illegal moves, Scholar's mate, castling and en passant on each.

About this test

Chess looks easy to draw and is hard to get right: a board that moves pieces is easy, a board that knows every rule is not. Each model was asked for a two-player game on one screen with every rule enforced, including check, checkmate, stalemate, castling, en passant and promotion. We played the same four sequences on each board, from a fresh page every time: two illegal moves (e2-e5 and the blocked Bf1-c4), Scholar's mate, kingside castling, and an en passant capture. On the boards that worked we also tried a check that must be answered, promotion and a ten-move stalemate. Look at what each board says when the game ends, and how it shows whose move it is.

Read more

The prompt, word for word

Create a playable chess game for two players on one screen: click a piece, then click where it should go.

It must follow all the rules of chess: only legal moves, check, checkmate, stalemate, castling, en passant and pawn promotion. Show whose turn it is and announce the result when the game ends.

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 5 of 5 2.0 min $0.075
2 GPT-6 Luna Best value 5 of 5 69 s $0.005
3 Claude Haiku 5.5 5 of 5 91 s $0.011
4 Mercury 3 5 of 5 101 s $0.015
5 DeepSeek V4.1 Flash 5 of 5 2.8 min $0.024
6 DeepSeek V4.1 Flash 1017 5 of 5 81 s $0.031
7 GPT-6.1 Terra 5 of 5 74 s $0.033
8 MiniMax M3.1 Flash (free route) 5 of 5 7.9 min $0.035
9 Muse Spark 1.3 5 of 5 3.9 min $0.052
10 Mistral Large 4 5 of 5 23 min $0.131
11 Gemini 4 Carbon 5 of 5 2.4 min $0.136
12 Kimi K3 5 of 5 3.7 min $0.267
13 Kimi K3.1 Pro 5 of 5 12 min $0.268
14 Qwen3.8 Max (0902) 5 of 5 20 min $0.281
15 Claude Opus 5.5 5 of 5 2.0 min $0.297
16 Grok 4.7 5 of 5 26 min3 tries $0.631
17 Claude Fable 5.1 5 of 5 2.9 min $0.859
18 Claude Fable 5.2 5 of 5 4.7 min $1.268
19 Muse Ember Pending 29 s $0.010
20 Gemini 3.8 Flash 0 of 5 110 s $0.062
21 Gemini 4 Argon Pending 40 s $0.065

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: GPT-6 Luna, 5 of 5 checks for $0.005. Our pick: GPT-6.1 Sol, 5 of 5 for $0.075. 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$10 Cost, every try (log scale) Checks passed GPT-6.1 Sol GPT-6 Luna

Best valueOur pickOther answers

Notes on each answer

1

GPT-6.1 Sol

Our pick 5 of 5

A plain slate-and-cream board with a side panel, a move list in proper notation and buttons to resign, offer a draw or claim one. It refused both illegal moves, announced "White wins! Checkmate." after Qxf7#, castled and took en passant correctly, and it also handled our check, promotion and stalemate probes ("Stalemate: no legal moves, and the king is not in check."). Nothing went wrong in our tests, and it was one of the quickest and cheapest working answers of the first ten.

2.0 min · 7,516 output tokens (790 thinking) · $0.075

2

GPT-6 Luna

5 of 5

A plain cream-and-green board with "White to move" above it, a New game button and the result below, but no move list, undo or flip. It refused both illegal moves, announced "White wins by checkmate." after Qxf7#, castled on both sides and took en passant, and our probes passed too: Black could not ignore Qh5+, promotion offers four pieces, and the ten-move stalemate ends with "Draw by stalemate." The board is laid out badly, though: a rank with no pieces on it shrinks to a strip, so at the start the four middle ranks are less than half the height of the others. The cheapest of the first twelve answers, at $0.0046 in 69 seconds.

69 s · 9,271 output tokens (3,047 thinking) · $0.005

Added 8 October 2026.
3

Claude Haiku 5.5

5 of 5

A brown wooden board with a dark side panel, a move list in standard notation ("4. Qxf7#") and resign and new-game buttons, but no undo or flip. All four sequences played correctly, and so did our probes: Black could not ignore Qh5+, promotion offers a choice of four pieces, and the ten-move stalemate ends with "Stalemate! The game is a draw." The status reads "White to move", adds "(in check!)" when needed, and shows "Checkmate! White wins." at the end. The cheapest of the first eleven answers, at about a cent.

91 s · 22,272 output tokens (13,195 thinking) · $0.011

Added 8 October 2026.
4

Mercury 3

5 of 5

A plain cream and tan board on a dark page, with "White to move." above it and a New Game button below; the chosen piece turns gold and dots show its moves. It refused both illegal moves and a black move on White's turn, played Scholar's mate to "Checkmate! White wins.", castled on both sides and took en passant, and our probes passed too: Black had to answer Qh5+, promotion offers four pieces and the ten-move stalemate ends "Stalemate! It's a draw." There is no move list, undo or flip. 101 seconds, $0.015.

101 s · 42,822 output tokens (36,160 thinking) · $0.015

Added 19 October 2026.
5

DeepSeek V4.1 Flash

5 of 5

A wooden board with a small side panel listing moves as "e2-e4" and a yellow "Check!" or green "Checkmate! White wins." under the turn line. All four sequences and our extra probes played correctly. It was the cheapest of the first ten answers, at about two cents.

2.8 min · 30,146 output tokens (21,274 thinking) · $0.024

6

DeepSeek V4.1 Flash 1017

5 of 5

A warm wooden board on a dark page, with a status pill ("White to move", "Black to move — Check!"), a king that glows red in check and a yellow "Checkmate — White wins!" line. It refused both illegal moves, played Scholar's mate to the end, castled on both sides and took en passant, and our probes passed too: Black had to answer Qh5+, promotion offers four pieces and the ten-move stalemate ends "Stalemate — it's a draw." It also handles the fifty-move, repetition and insufficient-material draws. There is no move list, undo or flip. 81 seconds, $0.031.

81 s · 26,025 output tokens (17,945 thinking) · $0.031

Added 19 October 2026.
7

GPT-6.1 Terra

5 of 5

A green and cream board with a status line, "New game" and "Claim draw". It refused both illegal moves, announced "Checkmate. White wins!", castled and took en passant correctly, and it also got our check, promotion and stalemate tries right. There is no move list, and empty rows are squashed, so the squares change height as pieces move. 74 seconds, $0.033.

74 s · 6,640 output tokens (1,124 thinking) · $0.033

Added 27 October 2026.
8

MiniMax M3.1 Flash (free route)

5 of 5

A dark green page with a cream and sage board, a card for each player, a status panel ("White to move", "Check — Black"), a move list in standard notation, and undo, flip, resign and draw buttons. It refused both illegal moves, played Scholar's mate to "White wins by checkmate", castled on both sides and took en passant, and our probes passed too: Black had to answer Qh5+, promotion offers four pieces and the ten-move stalemate ends "Draw by stalemate". The result banner covers the back rank, and the black pieces are hard to make out on the dark squares. One try, 7.9 minutes on OpenRouter's free route, so the time is not representative; $0.035 at list prices.

7.9 min · 57,569 output tokens (42,514 thinking) · $0.035

Added 19 October 2026.
9

Muse Spark 1.3

5 of 5

A clean board under a "White to move" bar, with the moves written in one line beneath it. Every sequence worked; checkmate and stalemate each dim the board behind a "Game Over" box, and promotion opens a piece picker. A correct, plain answer and the second cheapest of the first ten.

3.9 min · 12,265 output tokens (5,148 thinking) · $0.052

10

Mistral Large 4

5 of 5

A tidy board with a captured-pieces tray, a material count and a highlighted move list. Every rule worked, but once the game ends the status still reads "Black to move" above "Checkmate - White wins". It took 23 minutes.

23 min · 62,420 output tokens (43,580 thinking) · $0.131

11

Gemini 4 Carbon

5 of 5

A green and cream board with clear contrasting pieces and a turn banner. It rejected e2-e5 and the blocked f1-c4, announced White wins after Scholar’s mate, moved the rook when castling and removed the captured pawn on en passant. These were played through the board, from a fresh game for each sequence. 1 try, 2.4 min, $0.1358.

2.4 min · 13,558 output tokens (2,988 thinking) · $0.136

Added 11 November 2026.
12

Kimi K3

5 of 5

The most minimal answer: a board, a "White to move" line under the title and a New Game button, with no move list or undo. All four sequences played correctly, and check, promotion and stalemate ("Stalemate - it's a draw.") all work. It has the fewest extras of the working boards, but nothing in it is wrong.

3.7 min · 19,036 output tokens (11,867 thinking) · $0.267

13

Kimi K3.1 Pro

5 of 5

A green board in a dark frame with a status card, a dot for whose turn it is, the last move highlighted and a red glow on a king in check. All four sequences played correctly, and so did our check, promotion and stalemate tries. It keeps no move list and has only a new game button. The judge, which saw part of the code, doubted promotion; it works, with a named choice of four pieces. 12 minutes, $0.27.

12 min · 26,774 output tokens (18,255 thinking) · $0.268

Added 27 October 2026.
14

Qwen3.8 Max (0902)

5 of 5

"The Checkmate Room": a green felt board in a wooden frame, player cards with a "to move" tag, a score sheet and a gold "Checkmate 1-0" card at the end. Every rule we tried worked, but the h-file sticks out past the right edge of the frame, and it took almost 20 minutes to write.

20 min · 46,816 output tokens (33,879 thinking) · $0.281

15

Claude Opus 5.5

5 of 5

A brown wooden board with a side panel, an undo button and a move list ending in "1-0". Every sequence played correctly, and checkmate brings up a "Checkmate! White wins." box with a "View board" button so you can still see the final position. It is as complete as GPT-6.1 Sol and just as fast, but cost about four times as much.

2.0 min · 14,814 output tokens (4,435 thinking) · $0.297

16

Grok 4.7

5 of 5

A dark wooden board with "Black" and "White" bars above and below it, a status box ("White to move", "Checkmate - White wins"), a move list, undo and flip. All four sequences played correctly, and Scholar's mate dims the board behind a "Checkmate / White wins" box. It sent code only on its third try, after two tries with no code; that third try took over nine minutes and cost $0.29, and all three took 26 minutes and $0.63.

9.5 min · 47,955 output tokens (36,804 thinking) · $0.288 for this answer; 26 min and $0.631 over 3 tries

Took 3 tries: sent no code at 64,000 tokens; sent no code at 64,000 tokens; finished at 64,000 tokens.Billed for 1,335 input tokens, about 1,212 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.
17

Claude Fable 5.1

5 of 5

A compact board with the status in a bar above it, which turns red for "Black to move - Check!" and green for "Checkmate! White wins." All four sequences played correctly, and it offers promotion with a choice of four pieces. It is correct, with undo and a flip button, but it was the most expensive of the first eighteen answers by far.

2.9 min · 17,160 output tokens (7,534 thinking) · $0.859

18

Claude Fable 5.2

5 of 5

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.

4.7 min · 31,660 output tokens (18,071 thinking) · $1.268

Added 10 November 2026.
19

Muse Ember

29 s · 19,270 output tokens (11,682 thinking) · $0.010

Added 9 November 2026.
20

Gemini 3.8 Flash

0 of 5

The page shows a "White's Turn" bar and an empty dark square where the board should be. The script crashes as it loads ("Cannot access 'chess' before initialization"): the game draws the board before the game object exists, and the board never appears. The New Game button throws the same error, so no move can be played.

Failed every check.

110 s · 16,631 output tokens (6,790 thinking) · $0.062

21

Gemini 4 Argon

40 s · 6,518 output tokens · $0.065

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: Refuses illegal moves; Detects checkmate and announces the result; Castling works; En passant works; Shows whose turn it is. A partial pass counts as half.

Verdict

All but one of the boards handled the four sequences we played: illegal moves, Scholar's mate, castling and en passant. Of the first ten, eight also passed our check, promotion and stalemate tries; Grok 4.7, which only sent code on its third try, was not put through those. With correctness level, speed and cost decide. GPT-6.1 Sol is our pick: correct throughout, with a move list in standard notation and draw and resign buttons, in two minutes for $0.075. Claude Opus 5.5 is just as good and about as fast (two minutes) but cost $0.30. DeepSeek V4.1 Flash ($0.024) and Muse Spark 1.3 ($0.052) were the cheapest fully working boards until Claude Haiku 5.5 and GPT-6 Luna, both added on 8 October, played every test correctly for $0.011 and $0.0046; Luna's board squashes its empty rows. DeepSeek V4.1 Flash 1017, added on 19 October, also played every test correctly ($0.031), but keeps no move list. MiniMax M3.1 Flash, added the same day on OpenRouter's free route, passed every test too, with a move list, undo and draw buttons, and so did Mercury 3, added the same day, on a bare board ($0.015). GPT-6.1 Terra, added on 27 October, played every test correctly ($0.033), but keeps no move list and squashes its empty rows. Kimi K3.1 Pro, added the same day, did too, on a clear board with the last move marked, but took 12 minutes and cost $0.27. Claude Fable 5.2, added on 10 November, played every test correctly as well, with a move list, captured pieces, undo and a flip button, but at $1.27 it is the dearest of the first nineteen boards. Claude Fable 5.1 cost $0.86, while Qwen3.8 Max (20 minutes) and Mistral Large 4 (23 minutes) took far longer to get the same result.

Gemini 3.8 Flash is the one failure: it crashes as the page loads and never draws a board.

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