aider/docs/llms.md

# Aider can connect to most LLMs

[![connecting to many LLMs](/assets/llms.jpg)](https://aider.chat/assets/llms.jpg)

Aider works best with [GPT-4 Turbo](#openai) and [Claude 3 Opus](#anthropic),
as they are the very best models for editing code.
Aider also works quite well with [GPT-3.5](#openai).

To use aider with a *free* API provider, you can use [Groq's Llama 3 70B](#llama3)
which is comparable to GPT-3.5 in code editing performance.
Cohere also offers free API access to their [Command-R+ model](#cohere), which works with aider
as a *very basic* coding assistant.

Aider supports connecting to almost any LLM,
but it may not work well with some models depending on their capabilities.
For example, GPT-3.5 is just barely capable of reliably *editing code* to provide aider's
interactive "pair programming" style workflow.
So you should expect that models which are less capable than GPT-3.5 may struggle to perform well with aider.

## Configuring models for aider

- [OpenAI](#openai)
- [Anthropic](#anthropic)
- [Llama3](#llama3-70b)
- [Cohere](#cohere)
- [Azure](#azure)
- [OpenAI compatible APIs](#openai-compatible-apis)
- [Other LLMs](#other-llms)
- [Editing format](#editing-format)

## OpenAI

To work with OpenAI's models, you need to provide your
[OpenAI API key](https://help.openai.com/en/articles/4936850-where-do-i-find-my-secret-api-key)
either in the `OPENAI_API_KEY` environment variable or
via the `--openai-api-key` command line switch.

Aider has some built in shortcuts for the most popular OpenAI models and
has been tested and benchmarked to work well with them:

- **GPT-4 Turbo**: running `aider` with no arguments uses GPT-4 Turbo by default.
- **GPT-4 Turbo with Vision**: run `aider --4-turbo-vision` to use this vision capable model, allowing you to share images with GPT by adding them to the chat with `/add` or by naming them on the command line.
- **GPT-3.5 Turbo**: `aider --35-turbo`.

You can use `aider --model <model-name>` to use any other OpenAI model.
For example, if you want to use a specific version of GPT-4 Turbo
you could do `aider --model gpt-4-0125-preview`.

## Anthropic

To work with Anthropic's models, you need to provide your
[Anthropic API key](https://docs.anthropic.com/claude/reference/getting-started-with-the-api)
either in the `ANTHROPIC_API_KEY` environment variable or
via the `--anthropic-api-key` command line switch.

Aider has some built in shortcuts for the most popular Anthropic models and
has been tested and benchmarked to work well with them:

- **Claude 3 Opus**: `aider --opus`
- **Claude 3 Sonnet**: `aider --sonnet`

You can use `aider --model <model-name>` to use any other Anthropic model.
For example, if you want to use a specific version of Opus
you could do `aider --model claude-3-opus-20240229`.

## CROQ

Groq currently offers *free* API access.
They host the Llama 3 70B model, which works
well with aider and is comparable to GPT-3.5 in code editing performance.


### Llama3 70B

```
export GROQ_API_KEY=<your-key-goes-here>
aider --model groq/llama3-70b-8192
```

## Cohere

Cohere offers *free* API access with a reasonable rate limit.
Their Command-R+ works well with aider
as a *very basic* coding assistant.

**Command-R+**:

```
export COHERE_API_KEY=<your-key-goes-here>
aider --model command-r-plus
```

## Azure

Aider can be configured to connect to the OpenAI models on Azure.

```
export AZURE_API_KEY=your-key-goes-here
export AZURE_API_VERSION=2023-05-15
export AZURE_API_BASE=https://example-endpoint.openai.azure.com
aider --model azure/<your_deployment_name>
```

See the
[official Azure documentation on using OpenAI models](https://learn.microsoft.com/en-us/azure/cognitive-services/openai/chatgpt-quickstart?tabs=command-line&pivots=programming-language-python)
for more information on how to populate the above configuration values.

## OpenAI compatible APIs

If your LLM is accessible via an OpenAI compatible API endpoint,
you can use `--openai-api-base` to have aider connect to it.

You might need to use `--no-require-model-info` if aider doesn't
recognize the model you want to use.
For unknown models, aider won't have normal metadata available like
the context window size, token costs, etc.
Some minor functionality will be limited when using such models.

## Other LLMs

Aider uses the [litellm](https://docs.litellm.ai/docs/providers) package
to connect to hundreds of other models.
You can use `aider --model <model-name>` to use any supported model.

To explore the list of supported models you can run `aider --model <model-name>`
with a partial model name.
If the supplied name is not an exact match for a known model, aider will
return a list of possible matching models.
For example:

```
$ aider --model turbo

Unknown model turbo, did you mean one of these?
- gpt-4-turbo-preview
- gpt-4-turbo
- gpt-4-turbo-2024-04-09
- gpt-3.5-turbo
- gpt-3.5-turbo-0301
...
```

Depending on which model you access, you may need to provide an API key
or other configuration parameters by setting environment variables.
If any required variables are not set, aider will print an
error message listing which parameters are needed.

See the [list of providers supported by litellm](https://docs.litellm.ai/docs/providers)
for more details.


## Editing format

Aider uses 3 different "edit formats" to collect code edits from different LLMs:

- `whole` is a "whole file" editing format, where the model edits a file by returning a full new copy of the file with any changes included.
- `diff` is a more efficient diff style format, where the model specifies blocks of code to search and replace in order to made changes to files.
- `udiff` is the most efficient editing format, where the model returns unified diffs to apply changes to the file.

Different models work best with different editing formats.
Aider is configured to use the best edit format for all the popular OpenAI and Anthropic models.

For lesser known models aider will default to using the "whole" editing format.
If you would like to experiment with the more advanced formats, you can
use these switches: `--edit-format diff` or `--edit-format udiff`.