Skip to main content
affpapa.org
RU EN
skill CX Codex

compare-models — Claude Code skill: compare Replicate models

A skill that helps pick a Replicate model for the job by weighing cost, speed, quality and capabilities.

Install into Claude Code · Skills CLI (Vercel) All methods and platforms ↓ 
npx skills add replicate/skills --skill compare-models -a claude-code
Open on GitHub

What compare-models does

The skill walks the agent through a short workflow: build a shortlist from search or collections, fetch each model's schema, then compare inputs, outputs and supported features. Pricing comes from model metadata or the Replicate site, and quality gets checked by running the same prompts through every candidate.

Speed is read from metrics.predict_time on completed predictions. Official models are always warm; community models can cold-boot. Billing differs too — official models have predictable per-run pricing, while community models charge by GPU time. So after a few test runs, check the metrics field for the actual cost.

The skill also lays out the tradeoffs. Smaller distilled models are cheapest, at the cost of slower inference and lower quality. Official or schnell/turbo variants cut latency but raise the price per run. Models with ControlNet, masks or reference images give the most control, and the input setup gets more involved. For prompting technique it points to the neighbouring prompt-images and prompt-videos skills. If a community model fits and no official one does, the guidance suggests creating a deployment for consistent uptime.

What it helps with

  • shortlist Replicate models that fit a task
  • compare per-run cost of an official model against a community one
  • measure real inference time from metrics.predict_time
  • run identical prompts across models and compare the output
  • find a model with the inputs you need: masks, references, aspect ratios

Example prompts for the agent

  • compare Replicate image models by cost and speed
  • run the same prompt through three models and show the quality difference
  • which community models can replace the official one, and what will a run cost

How to install compare-models

1Skills CLI (Vercel)

npx skills add replicate/skills --skill compare-models -a claude-code

2Manually: put the files in place

~/.claude/skills/compare-models/SKILL.md

Copy the whole skill folder; for one project use .claude/skills/compare-models/

Step by step, with commands for each platform: How to install skills in Claude Code: five ways

Questions about compare-models

How do I install compare-models in Claude?

Run npx skills add replicate/skills --skill compare-models -a claude-code. The skill lands at ~/.claude/skills/compare-models/SKILL.md.

What does compare-models do?

It guides the agent through comparing Replicate models: shortlisting candidates, reading schemas, checking pricing, running test predictions and picking by cost, latency or quality.

How do official Replicate models differ from community models?

Official models stay warm, ship stable APIs and charge a predictable per-run price. Community models may cold-boot, need version pinning and bill by GPU-seconds.

Source text

SKILL.md · Apache-2.0 
---
name: compare-models
description: Compare Replicate models by cost, speed, quality, and capabilities.
---

## Docs

- Reference: <https://replicate.com/docs/llms.txt>
- OpenAPI schema: <https://api.replicate.com/openapi.json>
- MCP server: <https://mcp.replicate.com>
- Per-model docs: `https://replicate.com/{owner}/{model}/llms.txt`
- Set `Accept: text/markdown` when requesting docs pages for Markdown responses.

## Workflow

1. Search or browse collections to build a shortlist of candidate models.
2. Fetch each model's schema to compare inputs, outputs, and capabilities.
3. Check pricing from model metadata or the Replicate website.
4. Run a small batch of test predictions to compare output quality.
5. Pick the model that best fits your constraints (cost, latency, quality).

## What to compare

- **Speed**: Check `metrics.predict_time` on completed predictions for actual inference time. Official models are always warm. Community models can cold-boot.
- **Cost**: Official models have predictable per-run pricing. Community models charge by compute time (GPU-seconds). Run a few predictions and check the `metrics` field for actual cost.
- **Quality**: Run the same prompts through each model and compare outputs. Quality is subjective. Match it to your use case, not a leaderboard.
- **Capabilities**: Compare input schemas for supported features (reference images, masks, aspect ratios, streaming, multi-image input). Check output formats.

## Key tradeoffs

- Lowest cost: smaller/distilled models. Accept slower inference and lower quality.
- Lowest latency: official models or schnell/turbo variants. Accept higher cost per run.
- Highest quality: pro/max/quality variants. Accept slower inference and higher cost.
- Most control: models with ControlNet, masks, or reference images. Accept more complex input setup.

## Official vs community models

- Official models: always warm, stable APIs, predictable pricing, maintained by Replicate.
- Community models: may cold-boot, require version pinning, maintained by the author.
- If a community model meets your needs and an official model doesn't, consider creating a deployment for consistent uptime.

## Prompting guidance

For prompting techniques and task-specific guidance:

- Image generation and editing: see the [prompt-images](../prompt-images/SKILL.md) skill.
- Video generation: see the [prompt-videos](../prompt-videos/SKILL.md) skill.

© replicate · Apache-2.0 · replicate/skills

Similar skills

Everything for AI and ML →
huawei-cloud-ascend-models-deploy huaweicloud Deploys LLM, VL, Embedding and Rerank models on Huawei Cloud Ascend DevServer 910B and runs inference tests. CX Codex +1 424installs lambda-labs-gpu-cloud Orchestra Research This skill helps run Lambda Labs GPU instances for ML training and inference. Use it for SSH access, persistent storage, or multi-node clusters. CX Codex +1 939installs machine-learning-engineer VoltAgent Community A subagent for deploying, optimizing, and serving ML models in production. Use it for inference infrastructure, scaling, and model monitoring. 431installs scenario-model-comparison scenario-labs Runs one brief unchanged across several Scenario models and lines up cost, latency and quality in a single table. Useful when picking a default model. CX Codex +1 361installs repo-intake-and-plan lllllllama Scans a repository through its README and project files, then builds a small reproduction plan. Use it before running experiments. CX Codex +1 476,200installs huggingface-local-models huggingface Finds a GGUF repo on Hugging Face, matches a quant to your memory and starts it with llama.cpp. No guessing at filenames. CX Codex +1 1,900installs

From the same set7

prompt-images Prompting guide for image generation and editing models on Replicate. It helps structure detailed natural-language prompts instead of relying on keyword lists. CX Codex +1 2,600installs prompt-videos A skill with prompting techniques for video models on Replicate. Useful when you write video prompts or build generation features. CX Codex +1 2,400installs run-models Teaches coding agents to run AI models on Replicate via predictions, webhooks, and streaming. Useful for automating inference pipelines. CX Codex +1 2,200installs find-models This skill makes the agent query the Replicate API for current models instead of trusting names it remembers. Handy when you need a model for a task and want to avoid stale picks. CX Codex +1 2,300installs build-models Walks an agent through packaging a custom model with Cog and serving it locally. Useful when you have weights or a HuggingFace project and need a predict.py. CX Codex +1 1,800installs publish-models Guides Cog model releases to Replicate, with pre-release checks and CI/CD setup. Use it to publish a version manually or automate releases with GitHub Actions. CX Codex +1 1,700installs

Project sponsors

Companies that keep this analytics open
Gold sponsor
Nutra.Club

Impressions, clicks and code copies — open statistics. We keep the numbers public for the same reason NeBlask itself is public. Become a sponsor