release aidev SIG 3/5

Aider v0.76.0: thinking-token budgets, response notifications, Claude 3.7 Sonnet across providers

Aider adds a --thinking-tokens budget control and displays reasoning content returned by models. Claude 3.7 Sonnet is now supported on OpenRouter, Bedrock and Vertex AI and becomes the default model on OpenRouter; the old model shortcut flags (--4o, --opus) are deprecated in favour of --model.

PUBLISHED2025-03-10
OBSERVED2026-08-11
AGE1y
SOURCES1
  • --thinking-tokens CLI option controls token budget for models that support thinking
  • Reasoning content from LLMs that return it is now displayed; reasoning-tag handling improved
  • remove_reasoning setting deprecated, replaced by reasoning_tag
  • Desktop notifications when an LLM response is ready: --notifications, --notifications-command
  • Model support: QWQ 32B, GPT-4.5-preview, DeepSeek V3 free on OpenRouter (openrouter/deepseek/deepseek-chat:free), Claude 3.7 Sonnet on OpenRouter/Bedrock/Vertex AI, openrouter/anthropic/claude-3.7-sonnet and :beta
  • Default model on OpenRouter changed to Claude 3.7 Sonnet
  • Model shortcut args (--4o, --opus) deprecated in favour of --model
  • Switched to tree-sitter-language-pack
  • Fixes: EOF (Ctrl+D) handling, git staged-file read errors, git identity respecting global config, SSL verification control for model info requests, clearer empty-response warnings
  • Offers to install dependencies for Bedrock and Vertex AI models
featuremodel-supportdeprecationfix

COMMUNITY

No curated reactions recorded for this event. Facts and takes are kept in separate layers — community context is added by hand, never blended into the record above.