Datasets:

JOHNISKING
/

ai-coding-prompt-language

tool stringclasses 2 values	prompt_language stringclasses 5 values	run int64 1 5	input_tokens int64 47.7k 124k	cached_input_tokens int64 27k 95.5k	output_tokens int64 2.54k 9.39k	seconds int64 33 82	api_cost_usd float64 0.31 0.43 ⌀	agent_turns float64 2 3 ⌀	file_chars int64 6.45k 16.9k	file_lines int64 21 544	works bool 2 classes	ui_language stringclasses 5 values	reply_language stringclasses 5 values	non_ascii_letters_in_script int64 0 755
claude_code	en	1	123,666	95,504	8,537	73	0.42	3	14,992	487	true	en	en	0
claude_code	en	2	79,213	51,762	7,870	63	0.39	2	15,088	527	true	en	en	0
claude_code	en	3	122,899	95,138	8,120	71	0.4	3	14,371	496	true	en	en	0
claude_code	en	4	78,058	51,776	6,668	58	0.35	2	11,989	409	true	en	en	0
claude_code	en	5	78,474	51,776	7,290	64	0.37	2	13,059	448	true	en	en	0
codex	en	1	47,954	29,824	2,872	33	null	null	7,566	49	true	en	en	0
codex	en	2	48,815	37,504	3,803	41	null	null	10,045	117	true	en	en	0
codex	en	3	48,422	41,856	3,321	36	null	null	9,216	132	true	en	en	0
codex	en	4	48,600	34,688	3,476	39	null	null	9,761	118	true	en	en	0
codex	en	5	64,581	45,312	3,788	47	null	null	10,509	64	true	en	en	0
claude_code	ko	1	78,282	51,827	6,976	59	0.36	2	11,789	425	true	ko	ko	210
claude_code	ko	2	78,381	51,827	7,116	59	0.37	2	11,922	419	true	ko	ko	207
claude_code	ko	3	80,452	51,827	9,140	78	0.42	2	12,278	478	true	ko	ko	200
claude_code	ko	4	78,678	51,827	7,369	63	0.37	2	12,367	439	true	ko	ko	201
claude_code	ko	5	77,399	51,827	6,013	49	0.34	2	10,159	356	true	ko	ko	230
codex	ko	1	47,677	38,912	2,536	38	null	null	6,452	45	true	ko	ko	53
codex	ko	2	48,793	34,560	3,649	36	null	null	9,266	118	true	ko	ko	107
codex	ko	3	48,868	41,856	3,694	38	null	null	9,482	121	true	ko	ko	116
codex	ko	4	48,806	41,856	3,632	37	null	null	9,792	59	true	ko	ko	99
codex	ko	5	48,545	41,856	3,410	35	null	null	8,984	46	true	ko	ko	109
claude_code	ja	1	80,471	51,802	9,045	73	0.42	2	14,114	536	true	ja	ja	244
claude_code	ja	2	78,974	51,802	7,501	62	0.38	2	13,306	516	true	ja	ja	284
claude_code	ja	3	78,263	51,802	6,827	75	0.36	2	12,150	446	true	ja	ja	312
claude_code	ja	4	79,568	51,802	8,226	70	0.4	2	14,331	544	true	ja	ja	336
claude_code	ja	5	79,053	51,802	7,583	59	0.38	2	13,427	541	true	ja	ja	271
codex	ja	1	49,078	41,856	3,858	39	null	null	10,033	116	true	ja	ja	143
codex	ja	2	48,063	41,856	2,837	36	null	null	7,282	48	true	ja	ja	108
codex	ja	3	48,218	37,504	3,015	34	null	null	7,563	47	true	ja	ja	83
codex	ja	4	48,756	37,504	3,544	41	null	null	9,307	48	true	ja	ja	95
codex	ja	5	49,184	27,008	3,966	47	null	null	10,704	35	true	ja	ja	27
claude_code	es	1	80,598	51,803	9,212	82	0.42	2	13,484	448	true	es	es	0
claude_code	es	2	77,083	51,803	5,797	52	0.33	2	10,331	354	true	es	es	0
claude_code	es	3	76,570	51,803	5,008	49	0.31	2	9,215	293	true	es	es	0
claude_code	es	4	80,764	51,803	9,391	82	0.43	2	14,794	488	true	es	es	0
claude_code	es	5	80,134	51,803	8,841	82	0.41	2	13,672	492	true	es	es	0
codex	es	1	48,163	38,912	3,038	39	null	null	8,595	53	true	es	es	0
codex	es	2	48,192	38,912	3,058	38	null	null	8,545	52	true	es	es	0
codex	es	3	49,107	29,824	3,966	43	null	null	10,912	49	true	es	es	0
codex	es	4	47,805	27,008	2,666	34	null	null	7,125	51	true	es	es	0
codex	es	5	47,740	41,856	2,616	33	null	null	6,865	21	true	es	es	0
claude_code	uk	1	79,506	51,803	8,137	68	0.39	2	12,084	436	true	uk	uk	419
claude_code	uk	2	79,926	51,803	8,648	71	0.41	2	15,006	477	true	uk	uk	559
claude_code	uk	3	78,718	51,803	7,403	63	0.37	2	13,443	442	true	uk	uk	601
claude_code	uk	4	80,777	51,803	9,363	77	0.43	2	16,875	529	true	uk	uk	707
claude_code	uk	5	80,006	51,803	8,628	73	0.41	2	15,262	486	true	uk	uk	755
codex	uk	1	49,257	41,856	3,986	37	null	null	10,547	101	true	uk	uk	170
codex	uk	2	47,898	41,856	2,651	35	null	null	6,967	54	true	uk	uk	126
codex	uk	3	48,240	41,856	3,006	34	null	null	7,843	50	true	uk	uk	112
codex	uk	4	48,950	41,856	3,739	35	null	null	10,025	50	true	uk	uk	89
codex	uk	5	49,736	38,912	4,524	46	null	null	12,388	38	false	uk	uk	189

Coding in your own language: does the prompt language change what Claude Code and Codex cost?

The same brick-breaker task given to Claude Code and Codex in English, Korean, Japanese, Spanish and Ukrainian, five runs each (50 runs): input and output tokens, time, cost, whether the game ran, and the language of the game text and code comments.

Headline: asking in Korean, Japanese, Spanish or Ukrainian cost about the same as asking in English: over 5 runs per language, the median input tokens per task stayed within 1.2% of the English median in both tools. What changed was the game itself: its on-screen text, and Claude Code's code comments, followed the language of the prompt.

Source page, chart and method: https://tokensave.app/data/ai-coding-prompt-language
Related: Claude Code vs Codex same-task runs (https://huggingface.co/datasets/JOHNISKING/claude-code-vs-codex-tokens), web search by language (https://huggingface.co/datasets/JOHNISKING/claude-web-search-by-language)
DOI (Zenodo): 10.5281/zenodo.23454308 (the Zenodo record also has the raw logs, the generated games and the scripts)
Also on: Hugging Face · Kaggle
Measured: 2026-10-18 and 2026-10-19
License: CC BY 4.0

Key findings

The prompt language barely moves the bill for agentic coding. Median input tokens: Claude Code 79,213 in English vs 78,381 to 80,134 in the other four languages; Codex 48,600 vs 48,163 to 48,950. The prompt is about a hundred tokens; the agent's own instructions, tools and code make up the rest.
Output did not grow either: median 7,870 tokens in English vs 7,116 to 8,841 for Claude Code, 3,476 vs 3,038 to 3,739 for Codex, inside the run-to-run spread. A task cost $0.31 to $0.43 in any language at the API list prices Claude Code reports.
The run matters more than the language. The most expensive runs were English: 2 of 5 English Claude Code runs took an extra agent step (122,899 to 123,666 input tokens) and one English Codex run made an extra model call (64,581). None of the 20 non-English Claude Code runs took the extra step; with five runs per language this looks like chance.
In all 40 non-English runs the game's on-screen text was in the prompt's language. Claude Code wrote its code comments in that language in all 20 of its non-English runs and, in 3 of 5 Spanish runs, named variables and functions in Spanish (paleta, ladrillos, dibujarBola). Codex wrote 4 comments in its 20 non-English runs, all in English.
49 of 50 games loaded and responded to keyboard and mouse in a headless browser without script errors. The exception, a Codex run with the Ukrainian prompt, crashed on a canvas error (a particle effect drew a circle with a negative radius).
The language tax is a chat problem: in an agentic coding task it is lost in the agent's fixed overhead.

Files

ai-coding-prompt-language.csv: one row per run with tool, prompt_language, run, input_tokens, cached_input_tokens, output_tokens, seconds, api_cost_usd, agent_turns, file_chars, file_lines, works, ui_language, reply_language, non_ascii_letters_in_script
ai-coding-prompt-language.json: the same data plus the exact prompts

Results

Prompt language	Tool	Input tokens (median)	Input tokens (range)	Output tokens (median)	Time (median)	API cost (range)	Game works	Game text language
English	Claude Code	79,213	78,058 to 123,666	7,870	64 s	$0.35 to $0.42	5 / 5	English
English	Codex	48,600	47,954 to 64,581	3,476	39 s	–	5 / 5	English
Korean	Claude Code	78,381	77,399 to 80,452	7,116	59 s	$0.34 to $0.42	5 / 5	Korean
Korean	Codex	48,793	47,677 to 48,868	3,632	37 s	–	5 / 5	Korean
Japanese	Claude Code	79,053	78,263 to 80,471	7,583	70 s	$0.36 to $0.42	5 / 5	Japanese
Japanese	Codex	48,756	48,063 to 49,184	3,544	39 s	–	5 / 5	Japanese
Spanish	Claude Code	80,134	76,570 to 80,764	8,841	82 s	$0.31 to $0.43	5 / 5	Spanish
Spanish	Codex	48,163	47,740 to 49,107	3,038	38 s	–	5 / 5	Spanish
Ukrainian	Claude Code	79,926	78,718 to 80,777	8,628	71 s	$0.37 to $0.43	5 / 5	Ukrainian
Ukrainian	Codex	48,950	47,898 to 49,736	3,739	35 s	–	4 / 5	Ukrainian

Method

Both tools on one Windows laptop, effort medium, each run in its own empty folder, prompt sent through standard input.
Claude Code 2.1.295 (Opus 5.5): claude -p --permission-mode acceptEdits --output-format json, no MCP servers; input summed over every model incl. cache reads.
Codex CLI 0.162 (gpt-6-luna): codex exec --json -s workspace-write; tokens from turn.completed.
Runs 1-2 measured 2026-10-18 (English runs 1-2 are the same-task runs of the Claude Code vs Codex dataset); runs 3-5 for all five languages added 2026-10-19 with the same versions and settings, cycling through the languages in turn.
"Game works": loaded in headless Chromium with a canvas, took mouse clicks and arrow keys for about two seconds, no script errors.

Limitations

One small task, five runs per language and tool; longer conversations in another language would add more non-English text to the context.
Cost figures are list-price estimates; subscription plans are not billed per token.

Citation

TokenSave (2026). Coding in your own language: does the prompt language change what Claude Code and Codex cost? [Data set]. Zenodo. https://doi.org/10.5281/zenodo.23454308

Check your own Claude Code usage at tokensave.app.

Downloads last month: 30

Number of rows:

Total file size:

34.7 kB