Token cost by programming language: the same programs in 10 languages on Claude, Gemini and GPT
Abstract
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. Source page, chart and method: https://tokensave.app/data/token-cost-by-programming-language Related: token cost by data format (https://tokensave.app/data/token-cost-by-format) Measured: 2026-10-18 License: CC BY 4.0 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 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. 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. Gemini: Google AI Studio prompt token counter on Gemini 3.8 Flash, minus its one start token. 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]. https://tokensave.app/data/token-cost-by-programming-language