Aider is the odd one out in this pricing tracker because it doesn’t have pricing. It’s an open-source, Apache 2.0-licensed command-line tool you install with pip, and there’s no account to create, no seat to buy, and no subscription tier to pick. The entire “pricing model” is: the tool is free, and you pay whatever LLM provider you point it at.
That’s a real distinction from tools like Cursor or Windsurf, which bundle a model-usage allowance into a subscription price. Aider has no bundling at all. You point it at a model (Sonnet, or the equivalent for DeepSeek, OpenAI, Gemini, or a local model) and supply your own provider API key through an environment variable, and every request goes straight to that provider’s API, billed at that provider’s own rates. There’s no markup, no credit pool, no usage meter on Aider’s side — the cost you see on your Anthropic or OpenAI invoice is the cost, full stop.
What does that mean in practice? It depends entirely on which model you pick and how much you use it. Aider’s own documentation names Claude Sonnet, DeepSeek R1 and V3, and OpenAI’s o1, o3-mini, and GPT-4o as its best-performing, most-tested model options, but it isn’t locked to any of them — it connects to “almost any LLM,” including models running entirely on your own hardware, which costs nothing beyond your electricity bill. If you run Aider against a frontier model like Sonnet or GPT-4o for daily heavy coding work, expect a bill that lands somewhere in the same range as a Cursor Pro Plus or Ultra subscription, since you’re paying the same underlying API rates those tools pass through to you. If you run it against DeepSeek or a cheaper open-weight model, your cost drops substantially, often to a fraction of what a subscription-based competitor charges for comparable usage. There’s no single number to quote here, which is why our frontmatter marks this as “typically $10-200+/mo depending on model and usage” rather than a fixed figure — it’s a genuinely wide range, and it’s the user’s choice of model that determines where in that range you land, not anything Aider itself controls.
This pricing shape has real advantages if you’re cost-sensitive or already have API credits sitting somewhere. There’s no subscription to forget to cancel, no seat minimums for a team, and no vendor lock-in to a particular model provider — switch from Claude to DeepSeek mid-project with a flag change, and your cost structure changes instantly along with it. For teams, that also means Aider has no per-seat team plan, no centralized billing, and no admin console; if you want to track spend across a group of engineers, you’re managing that through whatever billing dashboard your LLM provider offers, not through Aider.
The tradeoff cuts the other way for anyone who wants predictable monthly costs. A $20/mo Cursor Pro subscription is a number you can budget around; API billing through Aider fluctuates with actual token consumption, and a long agentic session against a large codebase can spike a day’s spend well past what you’d expect from a subscription tool’s daily average. If your team values a flat, predictable line item for finance to approve, that’s a real argument for a subscription-based competitor even at a nominally higher headline price. If you’d rather pay exactly for what you use and pick your model freely without a vendor’s bundled allowance shaping that choice, Aider’s zero-markup, BYO-key model is close to as transparent as coding-tool pricing gets.
One more thing worth noting for anyone comparing this list at a glance: Aider genuinely has no changelog entries to report here, because there’s no pricing history to track. The tool has never charged for itself, and nothing about its cost structure has changed since it shipped — the only thing that moves is the API pricing set by whichever provider you connect, and those are covered separately in each provider’s own page in this tracker.