ModelPlan
Recommends which AI coding model, reasoning effort and coding plan to use, via a website, Skill, CLI and Claude Code and Codex plugins.
At a Glance
ModelPlan code (CLI, plugins, Skill) is released under the MIT license, free to use, copy, modify and distribute. The plan prices shown on modelplan.ai/plans belong to third-party vendors, not to ModelPlan.
Engagement
Available On
Listed Nov 2026
About ModelPlan
ModelPlan is an independent, open-source project that helps developers choose which AI coding model to run, at which reasoning effort, and on which coding plan. It publishes a leaderboard in which each model is ranked at its own best-value effort, and it offers the same data through a website, a read-only API, an agent Skill, a CLI and plugins for Claude Code and Codex. The code is MIT licensed, and the project says it is not affiliated with any model vendor.
What It Is
ModelPlan is a model and plan advisor for AI coding work. Most leaderboards compare models at their top setting, whereas ModelPlan marks a recommended (★) effort for each model and shows what one more step up adds in ability and costs in multiples. Ability, cost per reference coding task and task time are shown side by side. Models that are costlier and no stronger than an alternative are flagged as outclassed.
Where It Runs
The same published data reaches users through several surfaces. The Skill is installed with a single command and works in Claude Code, Codex, Cline, Amp and other agents. The Claude Code plugin adds slash commands and a status line. The Codex plugin adds a GPT effort dashboard. The modelplan CLI is a single static binary for macOS and Linux, and a macOS menu bar desktop app is available as a preview. The project also lists a read-only API, llms.txt and an ICS calendar of off-peak hours and quota resets.
Quota-Aware Advice
The Skill and CLI can read local Claude Code and Codex quota windows on the user's machine, and the project states these are never uploaded. Optional connectors cover other plans such as Z.ai, OpenCode Go, Kiro, Cursor, MiniMax and Copilot. Based on remaining quota and reset times, the tools suggest stepping up, staying at ★ or stepping down. The project states that it only advises and never switches the model or effort itself, and that there are no accounts or telemetry.
Plans and Data Sources
Plan Depot lists coding plans by maker and model, including quotas and covered models, and Burn Windows shows off-peak hours and quota resets. The About page marks plan recommendations as Beta, with exact plan-model-effort coverage still being completed. Ability data comes from public benchmarks such as Artificial Analysis, and cost is computed from official vendor prices. The methodology and open data are published in the repository.
Community Discussions
Be the first to start a conversation about ModelPlan
Share your experience with ModelPlan, ask questions, or help others learn from your insights.
Pricing
Open Source (MIT)
ModelPlan code (CLI, plugins, Skill) is released under the MIT license, free to use, copy, modify and distribute. The plan prices shown on modelplan.ai/plans belong to third-party vendors, not to ModelPlan.
- Code licensed under MIT
- Open data under per-file licenses (CC BY-SA 4.0 for new data; historical CC BY 4.0)
- Desktop official visuals under a separate reserved license, not MIT
- Skill, CLI, Claude Code and Codex plugins
Capabilities
Key Features
- Recommended best-value effort for each model
- Leaderboard ranked at each model's recommended effort
- Ability, cost and task time comparison
- Outclassed model detection and switch suggestions
- Local Claude Code and Codex quota reading
- Optional connectors for other coding plans
- Plan Depot with coding plan quotas and covered models
- Off-peak hours and quota reset calendar
- Claude Code plugin with status line and slash commands
- Codex plugin with GPT effort dashboard
- Agent Skill installable via npx
- modelplan CLI single binary
- Read-only API, llms.txt and ICS calendar
- Open data and public methodology
