ladder

Compare local and frontier models side-by-side on fixture-driven tasks.

10|5|Updated Jun 3, 2026

Search Tags:#onboarding #llm #tool-calling #local-model #frontier #ladder

Ready to run

npx skills add https://github.com/understudylabs/understudy-agent-tools --skill ladder-understudylabs

💬 Prompt for ChatGPT / Claude Chat▼

Please help me install this Agent Skill.
Skill: ladder
Source: https://github.com/understudylabs/understudy-agent-tools/tree/main/skills/ladder
Command: npx skills add https://github.com/understudylabs/understudy-agent-tools --skill ladder-understudylabs

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

Ladder provides a local-first, no-data demonstration that lets developers compare a small local model to a frontier model in a single UI, highlighting live reasoning, scoring, and tool-calling behavior.

Core Features & Use Cases

Local-vs-frontier side-by-side comparison with live scoring; VS mode; onboarding demo; no data uploads; fixture-driven tasks; headless and interactive usage.
Use Case: Onboard new users by showing the concrete difference between local and frontier models on representative tasks and scoring results.

Quick Start

Start the ladder by running uv run --with mlx-vlm --with mlx-lm python skills/ladder/serve.py and open http://127.0.0.1:8011/ladder.climb.html.

Frequently Asked Questions about ladder

High-intent search queries and answers about installing and using this skill.

FAQPage Schema

How do I compare local models to frontier models side-by-side?▼

Compare local models to frontier models side-by-side by running a local-first ladder demo that renders live reasoning, tool-calling behavior, and streaming scores against fixture-driven tasks in a single UI.

What is the best way to onboard new users on local vs frontier model differences?▼

Onboard new users on local vs frontier model differences by demonstrating concrete performance gaps on classify tasks and hard tool-calling tasks with live scoring, without requiring any data uploads.

How do I test tool-calling tasks with a local model safely?▼

Test tool-calling tasks with a local model by running fixture-driven tasks that include anti-shotgun safeguards, rendering tool calls and reasoning in a frontend UI to evaluate behavior safely.

Do I need to upload my own data to compare local and frontier models?▼

No, you do not need to upload your own data to compare local and frontier models because the demo operates entirely on local-first, fixture-driven tasks for safe and immediate evaluation.

Can I use a gateway option for billed frontier model runs during comparison?▼

Yes, you can use a gateway option for billed frontier model runs during comparison, allowing you to evaluate frontier performance alongside a small local model in the same interface.