aider

New Models

• Added support for new Gemini models including gemini-2.5-pro, gemini-2.5-flash, and gemini-2.5-pro-preview-06-05 with thinking tokens support.
• Updated model aliases: flash now points to gemini-2.5-flash and gemini now points to gemini-2.5-pro.
• Support for Responses API models like o1-pro, o3-pro.
• Added support for OpenAI o3-pro model across multiple providers.
• Enhanced thinking tokens support: can now be disabled by setting to 0, and improved help text with examples.
• Increased max tokens for Deepseek models to 65536 for better performance.
• Updated pricing for o3.

New Features & Enhancements

• Added --add-gitignore-files flag to enable adding files listed in .gitignore to Aider's editing scope, by omarcinkonis.
• Co-authored-by attribution is now enabled by default for commit messages.
• Enhanced commit message generation to use system prompt prefixes, by Luke Reeves.
• Added --commit-language option to specify the language for commit messages, by Kyosuke Takayama.
• Skip expensive file tracking operations when --skip-sanity-check-repo is enabled for better performance, by Makar Ivashko.
• Optimized chat history summarization performance, by jayeshthk.
• Added custom PostHog analytics configuration options with --analytics-posthog-host and --analytics-posthog-project-api-key flags, by Vasil Markoukin.

Repository Map

• Added MATLAB language support for repository maps, by Matthew Tofano.
• Added Clojure language support for repository maps, by Garrett Hopper.
• Improved kebab-case identifier recognition in repository maps for better code analysis.

Fixes & Small Improvements

• Improved GitHub Copilot token handling with better validation and error messages, by Vincent Taverna and Sebastian Estrella.

• Improved inline code rendering in Rich markdown output, by Vamsi Talupula.

• Fixed Vertex AI model name prefixes in settings, by Wietse Venema.

• Improved /read-only command to resolve literal paths correctly, by Matteo Landi.

• Fixed encoding issues in git diff output and LLM history logging.

• Fixed search block regex to accept optional closing tags when working with HTML content, by Mathis Beer.

• Auto-create parent directories for chat history files to prevent startup errors, by contributor.

• Ensure pip is available before package installation.

• Aider wrote 21% of the code in this release.

Model support & provider updates

• Added support for new Claude models including the Sonnet 4 and Opus 4 series (e.g., claude-sonnet-4-20250514, claude-opus-4-20250514) across various providers. The default sonnet and opus aliases were updated to these newer versions.
• Added support for the vertex_ai/gemini-2.5-flash-preview-05-20 model.
• Updated default OpenRouter models during onboarding to deepseek/deepseek-r1:free for the free tier and anthropic/claude-sonnet-4 for paid tiers.
• Automatically refresh GitHub Copilot tokens when used as OpenAI API keys, by Lih Chen.

Core functionality improvements

• Fixed an issue where files explicitly added via the command line were not correctly ignored if listed in .gitignore.
• Improved automatic commit messages by providing more context during their generation, by wangboxue.
• Fixed OpenRouter token cost calculation for improved accuracy.

Quality of life improvements

• Improved OpenRouter model metadata handling by introducing a local cache, increasing reliability and performance.
• Added shell tab completion for file path arguments (by saviour) and for --edit-format/--editor-edit-format options.
• The /settings command now displays detailed metadata for active main, editor, and weak models.
• Improved user language detection by correctly normalizing hyphenated language codes (e.g., en-US to en) and enhancing the validation of locale results.
• Prevented aider from instructing the LLM to reply in 'C' or 'POSIX' when these are detected as the system locale.
• Displayed a spinner with the model name when generating commit messages.
• Bumped configargparse to 1.7.1 as 1.7 was pulled.

• Added support for gemini-2.5-pro-preview-05-06 models.
• Added support for qwen3-235b models.
• Added repo-map support for OCaml and OCaml interface files, by Andrey Popp.
• Added a spinner animation while waiting for the LLM to start streaming its response.
• Updated the spinner animation to a Knight Rider style.
• Introduced --attribute-co-authored-by option to add co-author trailer to commit messages, by Andrew Grigorev.
• Updated Gemini model aliases (e.g., gemini, gemini-2.5-pro) to point to the 05-06 preview versions.
• Marked Gemini 2.5 Pro preview models as overeager by default.
• Commit message prompt specifies the user's language.
• Updated the default weak model for Gemini 2.5 Pro models to gemini/gemini-2.5-flash-preview-04-17.
• Corrected gemini-2.5-pro-exp-03-25 model settings to reflect its lack of support for thinking_budget.
• Ensured model-specific system prompt prefixes are placed on a new line before the main system prompt.
• Added tracking of total tokens sent and received, now included in benchmark statistics.
• Automatically fetch model parameters (context window, pricing) for OpenRouter models directly from their website, by Stefan Hladnik.
• Enabled support for thinking_tokens and reasoning_effort parameters for OpenRouter models.
• Improved cost calculation using litellm.completion_cost where available.
• Added model settings for openrouter/google/gemini-2.5-pro-preview-03-25.
• Added --disable-playwright flag to prevent Playwright installation prompts and usage, by Andrew Grigorev.
• The aider scrape command-line tool will now use Playwright for web scraping if it is available, by Jon Keys.
• Fixed linter command execution on Windows by adopting oslex for argument quoting, by Titusz Pan.
• Improved cross-platform display of shell commands by using oslex for robust argument quoting, by Titusz Pan.
• Improved /ask mode to instruct the LLM to elide unchanging code in its responses.
• Ensured web scraping in the GUI also respects Playwright availability and the --disable-playwright flag.
• Improved display of filenames in the prompt header using rich Text formatting.
• Enabled reasoning_effort for Gemini 2.5 Flash models.
• Added a --shell-completions argument to generate shell completion scripts (e.g., for bash, zsh).
• Explicit --attribute-author or --attribute-committer flags now override the default behavior when --attribute-co-authored-by is used, allowing finer control over commit attribution, by Andrew Grigorev.
• Fixed an issue where read-only status of files might not be preserved correctly by some commands (e.g. /drop after adding a read-only file).
• The aider-args utility (or python -m aider.args) now defaults to printing a sample YAML configuration if no arguments are provided.
• Displayed token count progress and the name of the file or identifier being processed during repo map updates.
• Extended the waiting spinner to also show for non-streaming responses and further enhanced its animation with console width clipping, cursor hiding, and a more continuous appearance.
• Dropped support for Python 3.9.
• Aider wrote 55% of the code in this release.

Full release notes: https://aider.chat/HISTORY.html

• Support for GPT 4.1, mini and nano.

• Improved support for using architect mode with Gemini 2.5 Pro.

• Add support for xai/grok-3-beta, xai/grok-3-mini-beta, openrouter/x-ai/grok-3-beta, openrouter/x-ai/grok-3-mini-beta, and openrouter/openrouter/optimus-alpha models.

• Added support for grok-3-fast-beta and grok-3-mini-fast-beta models.

• Added new patch edit format for OpenAI's GPT-4.1 model.

• Added new editor-diff, editor-whole, and editor-diff-fenced edit formats.

• Bugfix for automatically selecting the best edit format to use in architect mode.

• Add alias "grok3" for xai/grok-3-beta.

• Add alias "optimus" for openrouter/openrouter/optimus-alpha.

• Fix URL extraction from error messages.

• Allow adding files by full path even if a file with the same basename is already in the chat.

• Fix quoting of values containing '#' in the sample aider.conf.yml.

• Add support for Fireworks AI model 'deepseek-v3-0324', by Felix Lisczyk.

• Aider wrote 92% of the code in this release.