# What about token optimizers?

**Date:** 2026-09-23  
**Author:** Kees C. Bakker  
**Original:** https://keestalkstech.com/notes/what-about-token-optimizers/

---

I was looking into token optimizers for OpenCode and found benchmarks of popular tools: [Caveman](https://github.com/juliusbrussee/caveman), [Rust Token Killer](https://github.com/rtk-ai/rtk) and [Ponytail](https://github.com/DietrichGebert/ponytail).

[
        
                            
                    ![](https://blog.jetbrains.com/wp-content/uploads/2026/07/JB-social-BlogSocialShare-1280x720-1-5.png)
                
            
            
                Speaking to AI Agents like Cavemen Saves 65% of Tokens. We Test.
                A paired A/B benchmark of the token-compression skill Caveman on Claude Code, run on SkillsBench: does it actually save tokens, and does it degrade AI agent output quality? Advertised saving: 65%.
                                    blog.jetbrains.com
                            
        
    ](https://blog.jetbrains.com/ai/2026/07/speak-to-ai-agents-like-cavemen-tosave-tokens/)

[
        
                            
                    ![](https://blog.jetbrains.com/wp-content/uploads/2026/07/JB-social-BlogSocialShare-1280x720-1-7.png)
                
            
            
                rtk Claude Code Token Savings: A Skill Trial Benchmark
                Does "rtk" reduce Claude Code token usage? Part 2 of a series where we take public “token saving” add-ons for coding agents and run the same paired A/B benchmark against each of them. Part 1 was th
                                    blog.jetbrains.com
                            
        
    ](https://blog.jetbrains.com/ai/2026/07/rtk-claude-code-token-savings/)

[
        
                            
                    ![](https://blog.jetbrains.com/wp-content/uploads/2026/07/JB-social-BlogSocialShare-1280x720-1-10.png)
                
            
            
                Ponytail Skill for Claude Code: Does It Really Cut Tokens
                Part 3 of a series where we take public "token saver" add-ons for coding agents and run the same paired A/B benchmark against each of them. Part 1 was the caveman skill (advertised −65%, measured −8.5
                                    blog.jetbrains.com
                            
        
    ](https://blog.jetbrains.com/ai/2026/07/ponytail-skill-claude-tested/)

To me the conclusion is that `rtk` should work, but it does not. Caveman is not automatically activated, but works (-8%). Ponytail got some positive feedback (-10.3%), but it also won't auto activate; *it is the first tool in this series that clearly saved money*.