内容摘录
models
<p align="center">
<a href="https://crates.io/crates/modelsdev"><img src="https://img.shields.io/crates/v/modelsdev.svg?label=version" alt="Version"></a>
<a href="https://github.com/arimxyer/models/actions/workflows/ci.yml"><img src="https://github.com/arimxyer/models/actions/workflows/ci.yml/badge.svg" alt="CI"></a>
<a href="https://opensource.org/licenses/MIT"><img src="https://img.shields.io/badge/License-MIT-yellow.svg" alt="License: MIT"></a>
</p>
A fast CLI and TUI for browsing AI models, benchmarks, and coding agents.
**Models Tab**: Browse 2000+ models across 85+ providers from models.dev, categorized by type (Origin, Cloud, Inference, Gateway, Dev Tool)
**Agents Tab**: Track AI coding assistants (Claude Code, Aider, Cursor, etc.) with version detection and GitHub integration
**Agents CLI**: View changelogs, check release status, and compare versions for AI coding tools — agents status, agents claude, and more
**Benchmarks Tab**: Compare model performance across 15+ benchmarks from Artificial Analysis, with creator filtering by source, region, and type
<video src="https://github.com/user-attachments/assets/07c750f4-ca47-4f89-8a32-99e0be5004d8" controls width="100%"></video>
What's New
Agents CLI
**agents command** — view changelogs, check status, browse versions for AI coding tools directly from the terminal
**Status table** — see installed vs latest version, 24h release indicator, and release frequency at a glance
**Interactive picker** — fuzzy-select any version with --pick, view its changelog
**Dual entry point** — use as models agents or create an agents symlink for standalone usage
Recent Highlights
**Fast agents CLI** — concurrent GitHub fetching and version detection, on par with native Go tooling
**91% open weights match rate** — three-stage Jaro-Winkler pipeline for per-model open/closed detection
**~400 benchmark entries** from Artificial Analysis with creator filtering by region and type
**Optimized binary** — ~6MB release builds with strip, LTO, and panic=abort
Features
Models Tab
**CLI commands** for scripting and quick lookups
**Interactive TUI** for browsing and comparing models
**Provider categories** — filter and group providers by type (Origin, Cloud, Inference, Gateway, Dev Tool)
**Cross-provider search** to compare the same model across different providers
**Copy to clipboard** with a single keypress
**JSON output** for scripting and automation
Agents Tab
**Curated catalog** of 12+ AI coding assistants
**Version detection** — automatically detects installed agents
**GitHub integration** — stars, releases, changelogs, update availability
**Persistent cache** — instant startup with ETag-based conditional fetching
**Customizable tracking** — choose which agents to monitor
Benchmarks Tab
**~400 benchmark entries** from Artificial Analysis with quality, speed, and pricing scores
**Auto-updating** — data fetched fresh from CDN on every launch; GitHub Action refreshes source data every 30 minutes
**Creator sidebar** with 40+ creators — group by region or type with colored section headers
**Per-model open weights detection** — runtime matching against models.dev, with source filter toggle
**Quick-sort keys** — instantly sort by Intelligence, Date, or Speed
**Dynamic columns** — list columns adapt to show the most relevant benchmarks for the active sort
**Detail panel** — full benchmark breakdown with indexes, scores, performance, and pricing
Installation
Cargo (from crates.io)
Homebrew (macOS/Linux)
Scoop (Windows)
Pre-built binaries
Download the latest release for your platform from GitHub Releases.
Build from source
TUI Usage
Interactive Browser
Run models with no arguments to launch the interactive TUI:
!Models tab screenshot
Keybindings
**Global**
| Key | Action |
|-----|--------|
| ] / ` | Switch tabs (Models / Agents / Benchmarks) |
| ? | Show context-aware help |
| q | Quit |
**Navigation**
| Key | Action |
|-----|--------|
| j / ↓ | Move down |
| k / ↑ | Move up |
| g | Jump to first item |
| G | Jump to last item |
| Ctrl+d / PageDown | Page down |
| Ctrl+u / PageUp | Page up |
| Tab / Shift+Tab | Switch panels |
| ← / → | Switch panels |
**Search**
| Key | Action |
|-----|--------|
| / | Enter search mode |
| Enter / Esc | Exit search mode |
| Esc | Clear search (in normal mode) |
Models Tab
**Filters & Sort**
| Key | Action |
|-----|--------|
| s | Cycle sort (name → date → cost → context) |
| 1 | Toggle reasoning filter |
| 2 | Toggle tools filter |
| 3 | Toggle open weights filter |
| 4 | Cycle provider category filter (All → Origin → Cloud → Inference → Gateway → Tool) |
| 5 | Toggle category grouping |
**Copy & Open**
| Key | Action |
|-----|--------|
| c | Copy provider/model-id |
| C | Copy model-id only |
| o | Open provider docs in browser |
| D | Copy provider docs URL |
| A | Copy provider API URL |
Agents Tab
![Agents tab screenshot
**Filters & Sort**
| Key | Action |
|-----|--------|
| s | Cycle sort (name → updated → stars → status) |
| 1 | Toggle installed filter |
| 2 | Toggle CLI tools filter |
| 3 | Toggle open source filter |
**Actions**
| Key | Action |
|-----|--------|
| a | Open tracked agents picker |
| o | Open docs in browser |
| r | Open GitHub repo |
| c | Copy agent name |
Customizing Tracked Agents
By default, models tracks 4 popular agents: Claude Code, Codex, Gemini CLI, and OpenCode.
Press a in the Agents tab to open the picker and customize which agents you track. Your preferences are saved to ~/.config/models/config.toml.
You can also add custom agents not in the catalog:
See Custom Agents for the full reference.
Benchmarks Tab
!Benchmarks tab screenshot
**Quick Sort** (press again to toggle direction)
| Key | Action |
|-----|--------|
| 1 | Sort by Intelligence index |
| 2 | Sort by Release date |
| 3 | Sort by Speed (tok/s) |
**Filters & Grouping**
| Key | Action |
|-----|--------|
| 4 | Cycle source filter (All / Open / Closed) |
| 5 | Toggle region grouping |
| 6 | Toggle type grouping |
…