Aider: What's New in March 2025
A summary of all updates, new features, and bug fixes released for Aider during March 2025.
New Features in March 2025
- Added OpenRouter OAuth integration, offering to authenticate if no model or keys are provided.(v0.80.0)
- Enabled automatic selection of OpenRouter default model based on free/paid tier status when the API key is set but no model is specified.(v0.80.0)
- Prioritized specific Gemini and Vertex AI experimental models when no model is specified and corresponding keys are set.(v0.80.0)
- Added the `openrouter/deepseek-chat-v3-0324:free` model.(v0.80.0)
- Added `Ctrl-X Ctrl-E` keybinding to edit the current input buffer in an external editor.(v0.80.0)
- Added repomap support for the Scala language.(v0.80.0)
- Added support for the SOTA Gemini 2.5 Pro model.(v0.79.0)
- Added support for the DeepSeek V3 0324 model.(v0.79.0)
- Introduced a new /context command that automatically identifies which files need to be edited for a given request.(v0.79.0)
- Added /edit as an alias for the /editor command.(v0.79.0)
- Added an "overeager" mode for Claude 3.7 Sonnet models to help keep the model working within the requested scope.(v0.79.0)
- Added support for thinking tokens when using OpenRouter Sonnet 3.7.(v0.78.0)
- Introduced commands to switch between model types: /editor-model and /weak-model.(v0.78.0)
- Added the --git-commit-verify flag to control whether git commit hooks are bypassed.(v0.78.0)
- Added support for AWS_PROFILE when using Bedrock models, allowing use of AWS profiles for credentials.(v0.78.0)
- Added support for the openrouter/google/gemma-3-27b-it model.(v0.78.0)
- Enabled vi-like behavior when pressing enter in multiline-mode while in vi normal/navigation-mode.(v0.78.0)
- Added /think-tokens command to set thinking token budget with support for human-readable formats (8k, 10.5k, 0.5M).(v0.77.0)
- Added /reasoning-effort command to control model reasoning level.(v0.77.0)
- Enabled support for Cohere model cohere_chat/command-a-03-2025 and Gemini model gemini/gemma-3-27b-it.(v0.77.0)
- Added --auto-accept-architect flag to automatically accept changes from architect coder format without confirmation.(v0.77.0)
- Added ability to control token budget for reasoning models using the new `--thinking-tokens` CLI option.(v0.76.0)
- Enabled display of thinking/reasoning content returned by LLMs.(v0.76.0)
- Introduced user notifications when the LLM has completed its last request and is waiting for input via the `--notifications` flag.(v0.76.0)
- Added support for specifying a custom desktop notification command using `--notifications-command`.(v0.76.0)
- Added support for the QWQ 32B model.(v0.76.0)
- Added support for DeepSeek V3 free model on OpenRouter via `--model openrouter/deepseek/deepseek-chat:free`.(v0.76.0)
- Added support for Claude 3.7 Sonnet models across OpenRouter, Bedrock, and Vertex AI.(v0.76.0)
- Added support for the GPT-4.5-preview model.(v0.76.0)
- Added support for Claude 3.7 Sonnet:beta on OpenRouter.(v0.76.0)
- Added support for the `openrouter/anthropic/claude-3.7-sonnet` model.(v0.76.0)
- Added an offer to install necessary dependencies when using Bedrock and Vertex AI models.(v0.76.0)
- Added basic support for the Claude 3.7 Sonnet model (use --model sonnet).(v0.75.0)
- Added support for the openrouter/o3-mini-high model configuration.(v0.75.0)
- Added support for tree-sitter language packs.(v0.75.0)
- Added HCL (Terraform) syntax support.(v0.75.0)
- Enabled markdown generation for o1 & o3-mini models by sending a specific formatting string.(v0.75.0)
Bug Fixes in March 2025
- Fixed linting errors for filepaths containing shell metacharacters.(v0.80.0)
- Fixed a bug in the `/run` command that was preventing auto-testing.(v0.80.0)
- Fixed a bug preventing `UnboundLocalError` during git tree traversal.(v0.80.0)
- Handled `GitCommandNotFound` error if git is not installed or not in PATH.(v0.80.0)
- Handled `FileNotFoundError` if the current working directory is deleted while aider is running.(v0.80.0)
- Fixed autocompletion for /ask, /code, and /architect commands.(v0.78.0)
- Fixed a bug where the default model was incorrectly set by deprecated --shortcut switches even when explicitly specified on the command line.(v0.77.0)
- Fixed handling of JSONDecodeError when loading the model cache file.(v0.77.0)
- Fixed handling of GitCommandError when retrieving git user configuration.(v0.77.0)
- Fixed weak_model_name patterns to correctly match main model name patterns for certain models.(v0.76.0)
- Fixed handling of Git errors encountered when reading staged files.(v0.76.0)
- Fixed Git identity retrieval to properly respect global configuration settings.(v0.76.0)
- Fixed an issue where the continuation prompt ". " was incorrectly included for multiline input.(v0.75.0)
- Fixed a bug related to the /editor command.(v0.75.0)
Improvements in March 2025
- Validated user-configured color settings on startup and warned/disabled invalid ones.(v0.80.0)
- Added a warning at startup if `--stream` and `--cache-prompts` are used together due to potential cost estimate inaccuracy.(v0.80.0)
- Boosted repomap ranking for files whose path components match identifiers mentioned in the chat.(v0.80.0)
- Changed web scraping timeout from an error to a warning, allowing scraping to continue with potentially incomplete content.(v0.80.0)
- Left-aligned markdown headings in the terminal output.(v0.80.0)
- Updated the edit format to the new model's default when switching models with `/model`, if the user was using the old model's default format.(v0.80.0)
- Fixed completion menu current item color styling.(v0.80.0)
- Added model setting validation to ignore --reasoning-effort and --thinking-tokens if the model does not support them.(v0.78.0)
- Annotated which models support reasoning_effort and thinking_tokens settings in the model settings data.(v0.78.0)
- Improved code block rendering in markdown output with better padding using NoInsetMarkdown.(v0.78.0)
- Enhanced the --aiderignore argument to resolve both absolute and relative paths.(v0.78.0)
- Improved platform information handling to gracefully handle retrieval errors.(v0.78.0)
- Bumped dependencies to incorporate the litellm fix for Ollama.(v0.78.0)
- Expanded programming language support significantly by adopting tree-sitter-language-pack, adding 130 new languages with linter support and 20 new languages with repo-map support.(v0.77.0)
- The /think-tokens and /reasoning-effort commands now display current settings when called without arguments.(v0.77.0)
- Display of thinking token budget and reasoning effort is now available in model information.(v0.77.0)
- Changed the --thinking-tokens argument to accept string values using human-readable formats.(v0.77.0)
- Improved AutoCompleter to require 3 characters for autocompletion to reduce noise.(v0.77.0)
- The bare /drop command now preserves original read-only files provided via args.read.(v0.77.0)
- Added ignore_permission_denied option to the file watcher to prevent errors when accessing restricted files.(v0.77.0)
- Enhanced handling of reasoning tags to improve cleanup of model responses.(v0.76.0)
- Improved error handling for EOF (Ctrl+D) input during user prompts.(v0.76.0)
- Added a helper function to ensure hex color values always include the leading # prefix.(v0.76.0)
- Improved SSL verification control for requests fetching model information.(v0.76.0)
- Improved handling of empty LLM responses by providing clearer warning messages.(v0.76.0)
- Deprecated the `remove_reasoning` setting, replaced by `reasoning_tag`, and added a deprecation warning.(v0.76.0)
- Switched to `tree-sitter-language-pack` for tree sitter support.(v0.76.0)
- Updated the default model on OpenRouter to Claude 3.7 Sonnet.(v0.76.0)
- Deprecated model shortcut arguments (like `--4o`, `--opus`) in favor of using the explicit `--model` flag.(v0.76.0)
- Added build.gradle.kts to the list of special files recognized for Kotlin projects.(v0.75.0)
- Prevented multiple cache warming threads from running simultaneously.(v0.75.0)
All Releases in March 2025
v0.80.06 features5 fixesThis release introduces significant integration updates, including OpenRouter OAuth support and new model availability. Several critical bugs related to filepaths, git operations, and the /run command have been resolved. Additionally, the tool now provides better startup validation for configurations and improves handling of external dependencies like git.
v0.79.05 featuresThis release focuses heavily on expanding model compatibility by adding support for Gemini 2.5 Pro and DeepSeek V3 0324. New context management features, including the /context command, improve workflow efficiency by automatically identifying necessary files for edits.
v0.78.06 features1 fixThis release introduces new model configuration flexibility, including support for thinking tokens with OpenRouter Sonnet 3.7 and AWS_PROFILE support for Bedrock models. Key usability enhancements include new commands for switching model types and fixes for autocompletion across several core commands.
v0.77.04 features3 fixesThis release introduces powerful new configuration commands, /think-tokens and /reasoning-effort, allowing fine-grained control over model behavior and budget. It also brings massive expansion to supported programming languages via tree-sitter integration and fixes several critical bugs related to model configuration and file handling.
v0.76.011 features3 fixesThis release significantly enhances reasoning model support by adding token budget control and displaying reasoning output directly. Aider now offers user notifications when LLM responses are ready and introduces broad support for new models like QWQ 32B and Claude 3.7 Sonnet across various providers. Additionally, several Git and configuration handling bugs have been resolved.
v0.75.05 features2 fixesThis release introduces basic support for the Claude 3.7 Sonnet model and expands language support with HCL (Terraform) and tree-sitter pack integration. Several bugs were fixed, particularly concerning multiline input continuation prompts and the /editor command.