ModelPlan
ModelPlan is an independent, open-source project that helps developers choose which AI coding model to use, at what reasoning effort, and on which coding plan. It publishes checkable public data on ability, cost, task time, plans, quotas and resets, and exposes the recommendations on the web and inside developer tools.
At a Glance
- Software developers and engineering teams using AI coding agents
- Individual users of Claude Code, Codex, Cline, Amp and terminal-based coding tools
- Teams comparing AI coding subscriptions and provider quotas
- Researchers and technically sophisticated users who want reproducible public benchmark and pricing data
AI Tools by ModelPlan
(1)ModelPlan
AI Coding Model Selection Guide
Discussions
No discussions yet
Be the first to start a discussion about ModelPlan
Latest News
ModelPlan CLI v0.5.53 released, with Codex reader behavior updates, cockpit public-data details, Apple-silicon app packaging and refresh/history fixes.
ModelPlan CLI v0.5.52 released with reader request-budget controls, local-history recovery, richer quota/reset details and app/CLI contract updates.
Dock beta internal test candidate published as a universal macOS build.
ModelPlan CLI v0.5.51 released with model quota status lines and additional CI/release work.
Products & Services
Web leaderboard and decision platform covering model ability, reference-task cost, task time, recommended effort, model makers and coding plans.
Beta catalogue of coding plans by maker and model, with prices, quotas, covered models and efforts, off-peak hours and quota resets.
Single-binary macOS and Linux command-line client for lineups, effort ladders, quota status, local usage history and advice.
Open-source agent skill for Claude Code, Codex, Cline, Amp and other agents that answers which model and effort to use for a task.
Market Position
ModelPlan positions itself against conventional AI-model leaderboards and plan-comparison tools by evaluating each model at its own recommended effort and combining ability, reference-task cost and task time. It adds quota-aware local tooling, plan coverage, off-peak windows and reset information; the methodology identifies Artificial Analysis, LiveBench, CursorBench, Terminal-Bench, OpenRouter and models.dev as adjacent benchmark, catalogue or discovery sources.
Founding Story
The project was started to address a limitation in typical model leaderboards: they compare models at their maximum setting, even though moving from Low to Max effort can multiply cost and time for only a few additional ability points. ModelPlan's initial vision is to compare each model at its own best-value effort and make the formulas, sources and clients publicly checkable.
Business Model
Revenue Model
ModelPlan is presented as an independent open-source project. The website, Skill and clients are available without accounts or sign-in; the repository is MIT licensed and the project states that it accepts no paid placement. No paid ModelPlan subscription or usage charge is described in the reviewed sources.
Target Markets
- Software developers and engineering teams using AI coding agents
- Individual users of Claude Code, Codex, Cline, Amp and terminal-based coding tools
- Teams comparing AI coding subscriptions and provider quotas
- Researchers and technically sophisticated users who want reproducible public benchmark and pricing data
- Choosing an AI coding model and reasoning effort for a refactor or other coding task
- Balancing coding quality against API/task cost and runtime
- Selecting a coding subscription plan based on included models, quotas and effort coverage
- Deciding whether to step up or down in effort based on remaining local quota and reset time
- Monitoring Claude Code, Codex and other supported coding-tool usage locally
- Embedding model-and-plan recommendations in agent workflows and terminal tooling