Larger context windows and configurable reasoning levels for GitHub Copilot

[ebndev]

GitHub Copilot now supports larger context windows and configurable reasoning levels to help you tackle deeper, more complex work.

Leverage one-million-token context windows

A one-million-token context window means users can now work across larger codebases, longer documents, and complex multi-file projects without losing context. This capability is available now in VS Code, Copilot CLI, and the GitHub Copilot app, and is expanding to more surfaces soon.

Configurable reasoning levels

Configurable reasoning levels let you dial in the right balance of speed and depth and unlock extended thinking for your hardest architectural and debugging challenges. This capability is available in VS Code, Copilot CLI, and the GitHub Copilot app, and is expanding to more surfaces soon.

A note on AI credit consumption

Choosing a larger context window or higher reasoning level will consume more AI credits per interaction. We recommend using the default context window and reasoning level for everyday tasks, and reaching for extended context or higher reasoning when you're tackling complex, multi-file problems.

Try it today

Select one of the supported models in VS Code, Copilot CLI or the GitHub Copilot app to try these new capabilities today. To get more information about either of these features, review our model documentation.

🌟 Leave a comment!

Join the discussion and leave feedback in the comments below!

[roshhellwett]

1M context is useful for large refactors — no more manually feeding files one by one. Configurable reasoning is the right call too, no point burning credits on simple autocomplete.

Only real concern: credit cost will sneak up on you if you're not careful. A live usage indicator before sending would go a long way.

[pchung13]

There's no way to define an agent to take advantage of this though...

[jtabox]

Guess I'll take on my mantle of Negative Nancy and say that a 1M context window is pretty much not worth it for private users who'd rather not spend $100 for 5 sessions.

If 5 sessions now burn $15 worth of credits with the small context window then using the 3x context window is simply not worth it. Same with copilot for private users.

But hey, hope your corporate clients keep you afloat and don't dump you as soon as something better comes up. Trading loyalty for a quick, ephemeral profit has always been the wisest choice through history.

[kaio326]

Cool. But when are you going to admit that the new pricing model was a mistake? Are you waiting for everyone that doesn't have a corporate contract to abandon you and cancel the subscription? I'm warning you right now, if people try Claude code they aren't coming back to copilot even if you rollback the pricing model

[jtabox]

There's no point in warning or threatening tbh. I'm pretty sure they're not imbeciles, they knew and had calculated exactly how much the increase would be, and any company will run a metric ton of risk/reward analyses before taking such a drastic step. It's not like some middle-manager woke up one day and decided to implement a new price system, and it's not a 10-15% price increase.

The whole thing started some months ago, when they limited access to most models from Pro and left some for Pro+. People were mad then too, but it was fine because you still had access to decent models with Pro, you just didn't have the latest Opus and GPT5, whatever. I assume they ran that system for a while but realized that it's not worth it so the next step was this one, to kneecap Pro/Pro+ subs to essentially minimal use and autocomplete.

I really can't see how this was not the intended goal. I've been using GPT 5.4 and its mini variant the last few months, and literally every month I'd end up with 40-60% credits left unused, after having run multiple (15+) sessions. This month, same models, got to 75% used after 4 sessions, already at like 4th of the month. That's supposed to be $15 worth of credit. I had activated extra pay-per-use because I assumed I'll probably end up using the $15 towards the end of the month.

Yeah, no... If I just continue doing exactly what I did up to last month, I'll end up paying $100. How tf is this something a non-braindead, non-corporate sub can think that "eyyy, so happy, now I only pay for what I use \o/". Just deactivated pay-per-use for starters and trying out open-source alternatives.

I'll go, but there's no one who can convince me this wasn't the intended purpose.

[tanvishinde017]

Thanks for the update! A one-million-token context window sounds especially useful for large repositories and multi-file refactoring tasks.

The configurable reasoning levels are also a nice addition since not every task requires maximum depth, and having control over the speed-versus-quality tradeoff should help optimize AI credit usage.

One thing I'd love to see in the future is better visibility into estimated credit consumption before submitting a request, especially when using extended context or higher reasoning levels.

Looking forward to trying these capabilities across more Copilot surfaces.