✨ Anthropic's Claude Sonnet 4.6 is now generally available and rolling out in GitHub Copilot. Early testing shows ➡️ It excels on agentic coding ➡️ It is particularly successful in search operations Try it out in Visual Studio Code or Copilot CLI.

@AnthropicAI @code you guys fast 👀

@AnthropicAI @code where is it?

@AnthropicAI @code please support 1m context

@AnthropicAI @code Focusing on search operations suggests we are moving past autocomplete into understanding how a codebase connects. It becomes a navigator for complex systems.

@AnthropicAI @code Tell me it's 1x credit

@AnthropicAI @code copilot with sonnet is actually cracked for agentic stuff

@AnthropicAI @code FYI @Merkousha @AmirNapster @KuomarsRahimi @mehrandvd

@AnthropicAI @code Sonnet 4.6 rolling out to GitHub Copilot "Excels on agentic coding" and "successful in search operations" AI coding assistants are evolving from autocomplete to actual agents

@AnthropicAI @code Wow from 30x this morning back to 1x

@AnthropicAI @code awesomeeeee! been waiting for it for so long haha, you guys using it for your developers?

@AnthropicAI @code been testing this and the search improvements are legit, GitHub. cuts down so much time finding the right functions in larger codebases

@AnthropicAI @code Copilot keeps leveling up.

@AnthropicAI @code They upgraded my brain and announced it on GitHub. Nobody cc'd me.

@AnthropicAI @code Hit a rate limit every time if I enqueue prompts

@AnthropicAI @code How big is the comtext window in the copilot?

@AnthropicAI @code exciting developments in ai coding. it sets a new standard for efficiency.

@AnthropicAI @code Can all new models be automatically enabled, or have a setting for this?

@AnthropicAI @code has anyone thought about the permission scope on Copilot's agentic mode? an agent that excels at search + code write has a very different blast radius than a chat assistant. if it has prod commit access, a compromised workflow is a supply chain risk 💀

@AnthropicAI @code Sonnet 4.6 popped up in the picker for me in the middle of a VS 2026 session today, I was impressed with its work.

@AnthropicAI @code iSon

@AnthropicAI @code the search improvement is the real story here. 90% of agentic coding failures come from the model not finding the right context, not from bad generation. if copilot can actually navigate a 500k line codebase without hallucinating file paths thats a game changer

"excels on agentic coding" is the key claim to watch. agentic coding requires two things the model has historically been bad at: staying within scope and following multi-step instructions without drifting. if 4.6 actually improved on those, it changes what you can hand off without supervision. the model upgrade matters most when the workflow around it can verify the output — otherwise you're just generating mistakes faster.

@AnthropicAI @code Why is Opus fast mode 30x? It used to be 9x and it was already way too high.

@AnthropicAI @code Claude Sonnet 4.6 on agentic coding sounds like a game-changer for Copilot. Can’t wait to see how it handles complex workflows.

@AnthropicAI @code been waiting for this - the search improvements alone are gonna save me hours every week

@AnthropicAI @code Good stuff. Although I must say I've just switched to direct Claude integration via an API key, just b/c of the 128k ctx window size. :/

@AnthropicAI @code Why always vscode why not visual Studio. Whyyy?

@AnthropicAI @code

@AnthropicAI @code Is there any extension to import GitHub’s issues? Lol didn’t know that

@AnthropicAI @code agentic coding and search are great, but how's the hallucination rate?

@AnthropicAI @code when 5.3 codex??

@AnthropicAI @code I think this Claude Sonnet thing might actually help me write more

@AnthropicAI @code Exciting news! Can't wait to try out Claude Sonnet 4.6 in Copilot. The improvements in agentic coding and search operations sound promising! Let's see how it enhances our coding experience! 🚀