Hugging Face's logo

Datasets:
JOHNISKING
/
token-cost-by-programming-language

Modalities:
Tabular
Text
Formats:
csv
Languages:
English
Size:
< 1K
Tags:
code
programming-languages
tokenizer
llm-cost
claude
gemini
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
Dataset Viewer
Auto-converted to Parquet Duplicate
language
stringclasses
10 values
program
stringclasses
3 values
characters
int64
1.11k
5.1k
lines
int64
46
199
claude55_tokens
int64
549
2.22k
claude45_tokens
int64
448
1.86k
gemini_tokens
int64
440
1.82k
gpt_tokens
int64
378
1.54k
Python
inventory
1,620
56
627
522
523
431
Python
dijkstra
1,267
51
549
448
440
378
Python
total
2,887
107
1,176
970
963
809
Ruby
inventory
1,368
63
613
503
499
433
Ruby
dijkstra
1,110
46
568
458
445
386
Ruby
total
2,478
109
1,181
961
944
819
JavaScript
inventory
1,691
70
734
616
609
491
JavaScript
dijkstra
1,504
56
708
583
575
476
JavaScript
total
3,195
126
1,442
1,199
1,184
967
TypeScript
inventory
1,806
68
758
632
623
511
TypeScript
dijkstra
1,804
59
832
679
665
559
TypeScript
total
3,610
127
1,590
1,311
1,288
1,070
PHP
inventory
2,104
80
846
705
694
620
PHP
dijkstra
1,593
59
763
606
568
508
PHP
total
3,697
139
1,609
1,311
1,262
1,128
Rust
inventory
2,347
87
988
800
782
638
Rust
dijkstra
1,734
55
827
652
629
551
Rust
total
4,081
142
1,815
1,452
1,411
1,189
Java
inventory
2,659
88
941
780
771
620
Java
dijkstra
2,442
66
993
771
740
597
Java
total
5,101
154
1,934
1,551
1,511
1,217
C++
inventory
2,862
104
1,087
934
926
766
C++
dijkstra
1,931
68
846
692
691
577
C++
total
4,793
172
1,933
1,626
1,617
1,343
Go
inventory
2,193
103
1,057
849
818
651
Go
dijkstra
2,099
96
1,086
900
883
708
Go
total
4,292
199
2,143
1,749
1,701
1,359
C
inventory
3,135
115
1,388
1,152
1,126
940
C
dijkstra
1,708
62
831
707
694
596
C
total
4,843
177
2,219
1,859
1,820
1,536

Token cost by programming language: the same programs in 10 languages on Claude, Gemini and GPT

Two small programs (an inventory class and Dijkstra's shortest path) written in Python, Ruby, JavaScript, TypeScript, PHP, Rust, Java, C++, Go and C, with input tokens counted on Claude Opus/Sonnet 5.5, Claude Haiku 4.5, Gemini 3.x and GPT (o200k_base).

Headline: both programs together cost 809 GPT tokens in Python and 1,536 in C (+90%). Python and Ruby were the cheapest on every tokenizer; Go and C the most expensive. Claude 5.5 needed 43% to 59% more tokens than GPT for the same code.

Key findings

  • Python and Ruby are the cheapest: 809 and 819 GPT tokens. JavaScript needs +20%, Java +50%, Go +68% and C +90%.
  • The ranking holds on every tokenizer tested.
  • Claude 5.5 costs 43% to 59% more than GPT for the same code (largest gap: Java). Gemini needs 12% to 25% more than GPT.
  • Types add tokens: TypeScript needs 11% more GPT tokens than the same JavaScript.
  • File size is a rough guide only: Java has the most characters, but Go and C need more tokens.

Files

  • token-cost-by-programming-language.csv: one row per language and program (inventory, dijkstra and total) with language, program, characters, lines, claude55_tokens, claude45_tokens, gemini_tokens, gpt_tokens
  • token-cost-by-programming-language.json: the same data plus all 20 source files, so every count can be reproduced

Data (both programs together)

Language Characters Claude Opus/Sonnet 5.5 Claude Haiku 4.5 Gemini 3.x GPT (o200k) GPT vs Python
Python 2,887 1,176 970 963 809 ±0%
Ruby 2,478 1,181 961 944 819 +1%
JavaScript 3,195 1,442 1,199 1,184 967 +20%
TypeScript 3,610 1,590 1,311 1,288 1,070 +32%
PHP 3,697 1,609 1,311 1,262 1,128 +39%
Rust 4,081 1,815 1,452 1,411 1,189 +47%
Java 5,101 1,934 1,551 1,511 1,217 +50%
C++ 4,793 1,933 1,626 1,617 1,343 +66%
Go 4,292 2,143 1,749 1,701 1,359 +68%
C 4,843 2,219 1,859 1,820 1,536 +90%

Method

  1. Each program written idiomatically in each language with no comments. Every version was compiled or run and printed exactly the same output as the Python version.
  2. Claude: each file sent through Claude Code (claude -p) with a fixed one-line system prompt; input tokens minus the same run with a one-character message. Opus 5.5 and Sonnet 5.5 gave identical counts; a re-run gave identical counts.
  3. Gemini: Google AI Studio prompt token counter on Gemini 3.8 Flash, minus its one start token.
  4. GPT: tiktoken with o200k_base.

Limitations

  • Two small programs. Comments, long names or boilerplate in real projects can shift the numbers; the ranking is more stable than the exact percentages.
  • Coding style changes the count within a language.
  • Input tokens only; not a measure of how well each model writes each language.

Citation

TokenSave (2026). Token cost by programming language: the same programs in 10 languages on Claude, Gemini and GPT [Data set]. Zenodo. https://doi.org/10.5281/zenodo.23446563

Count tokens for your own code at tokensave.app.
Downloads last month
22
Number of rows:
30
Total file size:
56.6 kB