Agent-Skills.md

Agent Skills: All models from a provider

Look up current LLM model information, capabilities, pricing, context windows, and comparisons using models.dev data. Use when asked about AI models, model capabilities, pricing, or provider comparisons.

referenceID: michaelfisher1997/opencode-config/models-dev

Repository

michaelfisher1997/opencode-config
MichaelFisher1997

Install this agent skill to your local

pnpm dlx add-skill https://github.com/MichaelFisher1997/opencode-config/tree/HEAD/skills/models-dev

Skill Files

Browse the full folder contents for models-dev.

Download Skill

Loading file tree…

skills/models-dev/SKILL.md

Skill Metadata

Name
models-dev
Description
Look up current LLM model information, capabilities, pricing, context windows, and comparisons using models.dev data. Use when asked about AI models, model capabilities, pricing, or provider comparisons.

Models.dev Reference

Live LLM model data from models.dev. Covers all major providers with up-to-date capabilities, pricing, context windows, release dates, and more.

When to Use

Use this skill when the user asks about:

  • What models a provider offers (e.g., "What models does Anthropic have?")
  • Model capabilities (reasoning, tool calling, attachments, etc.)
  • Pricing comparisons between models
  • Context window sizes
  • Which models support a specific feature
  • Model release dates or knowledge cutoffs
  • Comparing two or more models
  • Finding the cheapest model that meets certain criteria
  • Open-weight models

Do NOT use for local codebase questions, git operations, or non-model topics.

Finding Latest/Flagship Models

When asked about "current," "latest," "flagship," "newest," or "state of the art" models, you must follow this discovery process:

  1. Query by provider first — Run list --provider <provider> for each major provider (anthropic, openai, google, deepseek, mistral, etc.) in parallel. This discovers all models including ones you may not know about.

  2. Identify the latest by release_date — From the results, find the model(s) with the most recent release_date per provider. Models.dev data is sorted by last_updated, but release_date is the authoritative field for when a model launched.

  3. Then compare — Once you've identified the latest models per provider, use compare to present them side-by-side if requested.

CRITICAL: Never assume you know model names. The model landscape moves fast. Always query providers directly rather than searching for model names you already know. For example, searching claude-opus-4 will miss claude-opus-4-6. Searching gpt-5 will miss gpt-5.4.

Major providers to check:

  • anthropic — Claude family
  • openai — GPT and o-series
  • google — Gemini family
  • deepseek — DeepSeek models
  • mistral — Mistral family
  • xai — Grok family
  • meta — Llama family

How to Run

All commands use bun and output JSON to stdout. Run from the skill directory:

bun /home/admin/.config/opencode/skills/models-dev/models.ts <command> [args]

Data is fetched live from https://models.dev/api.json on each run.

Commands

Search

Find models by name, provider, ID, or family:

bun models.ts search claude
bun models.ts search gpt-5
bun models.ts search gemini
bun models.ts search reasoning

List with Filters

List models with optional filters (filters combine with AND logic):

# All models from a provider
bun models.ts list --provider openai

# Models with reasoning capability
bun models.ts list --reasoning

# Models with tool calling
bun models.ts list --tool-call

# Open-weight models
bun models.ts list --open-weights

# Combine filters
bun models.ts list --provider anthropic --reasoning

# By capability name
bun models.ts list --capability tool_call

# By family
bun models.ts list --family gemini

Compare Models

Side-by-side comparison of 2+ models:

bun models.ts compare claude-opus-4 gpt-5
bun models.ts compare claude-opus-4 gpt-5 gemini-2.5-pro

Pricing

Find models by cost with optional filters:

# All models with pricing, sorted by input cost
bun models.ts pricing

# Models with 128k+ context
bun models.ts pricing --context 128000

# Cheapest models (input cost)
bun models.ts pricing --sort input --limit 10

# Cheapest by output cost
bun models.ts pricing --sort output --limit 10

# Cap input cost at $5/1M tokens
bun models.ts pricing --input-cost 5

# Combine filters
bun models.ts pricing --context 100000 --input-cost 10 --sort input

Providers

List all providers and their model counts:

bun models.ts providers

Output Format

All commands return JSON arrays. Each model object contains:

{
  "provider": "anthropic",
  "providerName": "Anthropic",
  "id": "claude-opus-4",
  "name": "Claude Opus 4",
  "family": "claude",
  "reasoning": true,
  "tool_call": true,
  "attachment": true,
  "temperature": true,
  "open_weights": false,
  "knowledge": "2025-04",
  "release_date": "2025-05-22",
  "last_updated": "2025-05-22",
  "modalities": { "input": ["text", "image"], "output": ["text"] },
  "cost": { "input": 15, "output": 75 },
  "limit": { "context": 200000, "output": 32000 }
}

Fields may be null if the data is unavailable for a given model.

Rules

  • NEVER fabricate model information — always query the script
  • If the script returns no results, say so honestly
  • The data reflects models.dev's current state; it may lag behind very recent announcements
  • If a model isn't found, try searching by partial name or provider
  • Cite the provider and model ID when giving answers
  • When answering "what's current/latest/flagship," never assume model names — always query providers directly (list --provider <name>) to discover the newest releases. Model names change quickly (e.g., claude-opus-4 → claude-opus-4-6, gpt-5 → gpt-5.4)
  • Use release_date (not last_updated) to determine which model is newest

Examples

User: "What's the cheapest model with 128k context?"

bun /home/admin/.config/opencode/skills/models-dev/models.ts pricing --context 128000 --sort input --limit 5

User: "Compare Claude Opus and GPT-5"

bun /home/admin/.config/opencode/skills/models-dev/models.ts compare claude-opus-4 gpt-5

User: "Which models support tool calling and have reasoning?"

bun /home/admin/.config/opencode/skills/models-dev/models.ts list --tool-call --reasoning

User: "What models does Google offer?"

bun /home/admin/.config/opencode/skills/models-dev/models.ts list --provider google

User: "What are the current flagship models?" or "What's the latest from each provider?"

# Step 1: Discover latest models per provider (run in parallel)
bun /home/admin/.config/opencode/skills/models-dev/models.ts list --provider anthropic
bun /home/admin/.config/opencode/skills/models-dev/models.ts list --provider openai
bun /home/admin/.config/opencode/skills/models-dev/models.ts list --provider google
bun /home/admin/.config/opencode/skills/models-dev/models.ts list --provider deepseek
# Step 2: Identify models with the most recent release_date per provider
# Step 3: Compare them side-by-side if needed
bun /home/admin/.config/opencode/skills/models-dev/models.ts compare claude-opus-4-6 gpt-5.4 gemini-2.5-pro

Keywords

LLM, models, AI, pricing, context window, capabilities, reasoning, tool calling, provider, OpenAI, Anthropic, Google, Gemini, Claude, GPT, Mistral, DeepSeek, models.dev, model comparison