---
title: "Discussion on \"Same model, 68% to 88% fewer tokens: Antigravity CLI vs Claude Code vs Codex, measured\" | Hashnode"
description: "Discussion on \"Same model, 68% to 88% fewer tokens: Antigravity CLI vs Claude Code vs Codex, measured\". Update (Oct 23): I added more game runs, so every tool now has 5 (52 runs in total). The same Opus 5.5 now reads 71% to 89% fewer input tokens in Antigravity than in Claude Code (was 68% to 88%), and"
url: "https://hashnode.com/posts/same-model-68-to-88-fewer-tokens-antigravity-cli-vs-claude-code-vs-codex-measured/6ada2b62b3622264664752ac"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [Same model, 68% to 88% fewer tokens: Antigravity CLI vs Claude Code vs Codex, measured](https://tokensave.hashnode.dev/same-model-68-to-88-fewer-tokens-antigravity-cli-vs-claude-code-vs-codex-measured?utm_source=hashnode&utm_medium=feed)

Update (Oct 23): I added more game runs, so every tool now has 5 (52 runs in total). The same Opus 5.5 now reads 71% to 89% fewer input tokens in Antigravity than in Claude Code (was 68% to 88%), and

Oct 22 · 5 min read · tokensave.hashnode.dev  
Read the article

-   [#ai](https://hashnode.com/tag/ai)
-   [#claude-code](https://hashnode.com/tag/claude-code)
-   [#gemini-cli](https://hashnode.com/tag/gemini-cli)

## Written by

[Jaehyun Cho](https://hashnode.com/@jonhisking)

@jonhisking · Seoul, South Korea

  
[View profile](https://hashnode.com/@jonhisking)

Tagline Indie developer from Korea · building TokenSave, a free AI token & API cost calculator in 34 languages

I'm an indie developer from South Korea. I build TokenSave (tokensave.app), a free, private calculator for AI tokens and API costs. It runs entirely in your browser and supports 34 languages. I like measuring how different languages tokenize, and how much that costs.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.
