Aider pricing: the software is free; API costs depend on model settings

Aider, the open-source AI coding assistant that runs in the terminal, costs 0 yen. It is Apache-2.0 software, and aider.chat lists no plans, tiers, seats, or hosted service. So “Aider pricing” in practice means the usage-based fee for the model API called through Aider. The figure to budget for is tokens per session × your provider’s rate. The setting most likely to inflate that figure is Aider’s default for an unknown model ID. It takes two lines to fix; the details are below.

One clarification first: search results also include another company called “Aider,” which sells accounting close software for firms. Its published monthly prices are for that product and are unrelated to this CLI. This page covers only the CLI from aider.chat, distributed as the aider-chat package from the Aider-AI/aider repository.

What you pay for Aider, and what you don’t

Aider’s FAQ states that the tool is published as open source under Apache 2.0. There is no pricing or plan information on the homepage, FAQ, installation guide, or LLM connection pages (verified October 1, 2026). The bill is all generated beyond the endpoint.

Check the update status before standardizing on the tool across a team. The repository is not archived and has 49,309 stars, but the last commit to main was on May 22, 2026 (GitHub API, October 1, 2026). The two official sources disagree on the “latest release”: PyPI lists 0.86.2, published February 12, 2026, while GitHub Releases stops at v0.86.0, dated August 9, 2025. pip install installs the PyPI version, so use that as your reference. For Python, the requirement in the 0.86.2 package metadata, >=3.10,<3.13, is what pip actually enforces, so use it rather than the description on the installation instructions page.

The setting that most affects API costs

This is the part pricing tables don’t show. Aider chooses the edit format, repo map, and cache behavior by comparing the model ID with a built-in table. The ID used for the comparison is the exact string written after --model, including its prefix. If nothing matches, the defaults are edit_format: whole, use_repo_map: false, cache_control: false, and weak_model_name: null, as stated on the advanced model settings page.

In the release 0.86.2 source, lookup works in two stages. First it checks for an exact match against 313 fully qualified names, such as gemini/gemini-3-pro-preview and openai/gpt-5.2. Only if there is no match does it try substring rules, including sonnet-4-5, haiku-4-5, and gpt-4.1. To connect to an OpenAI-compatible gateway, the model ID must be passed as openai/<id>, but all 40 entries with the openai/ prefix in the table are OpenAI’s own model names. As a result, openai/claude-sonnet-4-6 matches neither stage, and Aider runs in whole-file editing, without a repo map. With whole-file editing, the model outputs each edited file in full on every turn. Output tokens are the expensive part of the bill.

Two lines fix it. Aider reads this file from your home directory, the git root, or the current directory. The key is the full name passed to --model.

# name は --model に渡す文字列と完全一致させる（接頭辞 openai/ も含む）
- name: openai/claude-sonnet-4-6
  edit_format: diff
  use_repo_map: true

For a one-off run, you can pass --edit-format diff when launching. By contrast, openai/claude-haiku-4-5 matches haiku-4-5 by substring and automatically enables diff and the repo map. This is a naming coincidence that can change with any Aider release, so check each time instead of relying on it.

There is another easy-to-miss effect. Aider calls the API in three roles: the main model, an editor model used in --architect mode, and a weak model used for commit messages and history summaries (options reference). If the ID is unrecognized, the source uses the main model for both the weak and editor roles, so all three roles are billed at the highest rate. Specify a cheaper ID with --weak-model to avoid this (the same openai/ prefix is also required).

A further effect is that an unrecognized ID also hides the cost display. Without input_cost_per_token, Aider shows the token count but no cost line, as the warnings page says: it “assumes the model is free.” This metadata file restores the line:

{
  "openai/claude-sonnet-4-6": {
    "input_cost_per_token": 0.0000021000,
    "output_cost_per_token": 0.0000105000,
    "litellm_provider": "openai",
    "mode": "chat"
  }
}

Even with metadata, Aider’s amount is its own calculation from reported token counts. Cache multipliers for a specific provider are also hard-coded, so use the amount as an estimate and compare it with your provider’s usage record.

Estimate per session

The following is token arithmetic, not a measured task cost or invoice ceiling. Assume one session sends 200,000 uncached input tokens and receives 12,000 output tokens in diff format. The whole-file column assumes that the same work produces 40,000 output tokens because edited files are returned in full. This is an illustrative figure, not a measurement. Rates are the current prices in the Kunavo catalog per million tokens.

With this assumption, the edit format alone changes the cost of the same work on Claude Sonnet 4.6 from $0.546 to $0.840. Fix the setting before comparing provider rates. Commit message and summary traffic (assume 30,000 input and 2,000 output tokens) costs Claude Sonnet 4.6 $0.084 if it stays on the main model, and Claude Haiku 4.5 $0.028 if you specify a weak model.

Multiply by the number of sessions per day to calculate a budget. Kunavo’s catalog price is a billing floor, not a ceiling. If the upstream provider reports a cost, the charge is the greater of the catalog price and the upstream cost multiplied by the applicable markup. Cache pricing and external tools are not included in this example.

“The cheapest API” is two questions

Aider has the only task-cost reference for this tool that can be checked again: the polyglot benchmark has each model solve 225 Exercism exercises across six languages under the same conditions and records the cost.

Values read from the leaderboard data file. Two caveats matter. The latest row is from October 3, 2025, so later models have not been measured here. Also, the cost column has no pricing date and shows what that run cost, not a rate available today. These models are cited as benchmark evidence; DeepSeek is not in the Kunavo catalog. The less expensive model in the estimate table above is Claude Haiku 4.5.

Here is the useful conclusion. o3-pro cost 5 times as much as gpt-5(high) and scored lower, so a high rate does not guarantee completion. A low rate does not guarantee the lowest total either: a cheap model that needs 3 retries can cost more than an expensive model that finishes in one try. Consider the cheapest model that completes the task with an acceptable review effort, then check the amount actually recorded in your account for that task. Aider’s /tokens measures “the number of tokens in the current chat context,” as defined in the command reference; it is not the total billed to date.

Choosing between the routes

Two limits apply to the gateway route. A gateway without its own Aider module connects through the compatible route documented on the OpenAI-compatible APIs page: OPENAI_API_BASE, OPENAI_API_KEY, and the required openai/ prefix in the model ID. There is no documented Anthropic-format base URL override, and the string ANTHROPIC_API_BASE does not appear in the 0.86.2 source. Also, do not count on a prompt-cache discount on this route: the cache page limits support to certain provider families, requires the explicit --cache-prompts option, and says cache statistics and costs are unavailable while streaming. This page has not tested whether cache works on the compatible route.

Using Aider with Kunavo and how to pay

Kunavo has a setup guide for this client. It is a reference for the published configuration, not a compatibility test. Aider has not been run against Kunavo’s endpoint, and all the information on this page comes from Aider’s released source code and official documentation. Keep your current working route, run a bounded task, and check the charge recorded in your account.

Payment is based on prepaid top-ups, with no monthly fee. The minimum top-up is $10 (the minimum amount needed to create a balance, not a task fee or subscription), and checkout is handled by Stripe. From Japan, you can use cards (Visa, Mastercard, American Express, JCB), Apple Pay, Google Pay, and Link. For details, see billing information; you can create an account here. To compare clients, the English Aider pricing page (the detailed version this page is based on) and Cline pricing may also be useful.

Frequently asked questions

Aider’s repository, package, documentation, and leaderboard data were checked on October 1, 2026. Behavior descriptions are based on the v0.86.2 release source. Kunavo rates are from the live catalog, and all monetary examples on this page are illustrative token calculations, not measured task costs.