GitHub Copilot is the AI coding assistant built into GitHub. The confusing part used to be which plan to pick. Now it’s that the plans are really a meter. Since June 1, 2026, every plan converts into a pile of AI credits, and how far those credits go depends on what you ask Copilot to do.
Everything below is accurate as of this morning, and AI pricing changes constantly. The specific numbers will drift. The way to think about the decision holds up longer.
What the money buys
One credit is a penny. Each plan comes with a monthly pile of them:
Free ($0): 2,000 inline code suggestions a month and a small, unspecified amount of chat and agent use.
Pro ($10 a month): unlimited suggestions and $15 in credits ($10 base plus a $5 bonus GitHub calls a flex allotment). Adds code review and handing tasks to Copilot to run in the cloud.
Pro+ ($39 a month): $70 in credits, plus handing work off to outside coding agents like Claude and OpenAI’s Codex.
Max ($100 a month): $200 in credits.
All four list the same models. As of October 28, there’s also Copilot Fusion. Instead of picking a model, you pick an effort level (Standard, High or Max) and Copilot picks the models, sometimes using a cheap one to draft and a stronger one to check. You pay each model’s normal rate, and for now paid plans get 30% off when using Fusion.
What it doesn’t buy
Predictability, mostly. Autocomplete suggestions don’t touch your credits, but chat and agents do, and agents are hungry. One review clocked a 40-minute agent session on Pro at about $12, which is most of the monthly allowance in one sitting.
Team stuff (pooled usage, admin controls, legal protection) lives on the per-person business plans.
The math
If you mostly use Copilot for autocomplete, Free or Pro covers you, because suggestions are free. If you run agents that work for half an hour at a time, Pro’s $15 might last about one long afternoon. That’s where Pro+ starts making sense: roughly $70 of usage for $39.
GitHub’s pitch for Fusion is that its Max level scored within two points of Claude Code on Opus 5.5 at about half the cost per task. That’s GitHub’s own comparison, so I’d treat it as a sales number until someone else repeats it.
The test that outlasts the prices: do you run out before the month does? If you never do, you’re on the right plan. If you’re out by the 10th, the next tier up is probably cheaper than buying extra credits a penny at a time.

