GitHub Copilot

What it actually does

Copilot writes code alongside you inside VS Code, JetBrains, Visual Studio and Neovim. You get inline completions as you type, a chat window for debugging and review, code translation between languages, workspace-aware context, and pull request summaries. It supports more than 20 languages and it is the most widely adopted assistant of its kind.

The completions are the part people actually rely on. Users credit it with cutting the time spent on boilerplate and routine work, and its G2 rating of 4.4 out of 5 across 397 reviews includes 70% five-star. Two cautions from the con list are worth carrying: it sometimes suggests outdated patterns, and it makes an easy crutch if you stop reading what it produces. Code is transmitted to servers for processing, which some teams cannot accept regardless of the productivity argument.

What it costs

Individual is $10 a month, Business is $19 per seat, and Pro+ is $39. Verified students, teachers, and maintainers of popular open-source projects get it free. Those headline prices did not rise, and code completions plus Next Edit Suggestions remain free and unlimited.

Now the part that matters more than the sticker price. On 1 June 2026 the flat "premium request" model was replaced by metered GitHub AI Credits, where 1 credit equals $0.01 and credits are consumed per token. That applied to all 4.7 million paid subscribers. GitHub announced it in late April 2026 and shipped it six weeks later. This was the second billing change in a year: June 2025 had already added $0.04 per extra premium request to plans that previously felt unlimited.

So the honest budget is not $10. It is $10 plus whatever your agent sessions burn, and that number is yours to discover.

Where it falls down

The metered model hits agentic use hardest, which is the exact workflow GitHub spent years promoting. Community reports document projected cost increases of 10x to 50x for developers running multi-step agent sessions. Tech Times covered the backlash on 1 June 2026. A $10 Pro plan's 1,000 credits can be gone in under two weeks because one multi-file agent session burns 30 to 40 credits. David Ramel wrote in Visual Studio Magazine on 4 June 2026: "I was shocked when a thorny problem with an AI agent in Visual Studio Code used most of my monthly Copilot Pro credits on day 1 of the new usage-based billing model."

The old safety net went with it. When credits run out, Copilot no longer falls back to a cheaper model. Premium features stop until you buy more credits or the cycle resets, so the failure mode is your work halting mid-task rather than degrading.

Annual subscribers got the worst of it. Model multipliers were raised mid-term, with Claude Opus 4.7 going from 7.5x to 27x on 1 June 2026. You prepaid for a year at one rate and the rate changed underneath you. Users also report being blocked at the premium-request limit even after raising their spending budget, documented in GitHub community discussion #166810, which means the spend controls do not always do what they claim.

Who should buy it

Most working developers, still. At $10 with free unlimited completions, this remains the most sensible default for people who mainly want good autocomplete and an occasional chat. Teams at $19 a seat get the policy and license controls they need. Students and open-source maintainers should take the free tier without hesitating.

Do not buy the $10 Pro plan expecting to run agents on it. If agentic coding is your main workflow, price out a month of real usage before committing, and assume the metered bill is the true cost. Do not prepay annually either, because the Opus multiplier change proved an annual commitment does not lock your rate. And if your employer forbids sending source code to third-party servers, this conversation ends there.

Our 3.5 is a good tool marked down for billing conduct, not for what it produces.