Gemini Code Assist Standard and GitHub Copilot Business both cost $19 per user per month. Not roughly, exactly. Google prices Standard at $0.026027397 an hour on a 12-month commitment, which is $228 a year, and GitHub prices Copilot Business at $19 a seat. The comparison almost every article still runs, that Google gives you a generous free tier and GitHub does not, describes a product Google stopped serving on 18 June 2026. Once you take the free tiers out, price is not the deciding factor between these two. Where your code lives is.
This matters because the free tier is what most of the published comparisons are built on. You will still find articles explaining that Gemini Code Assist gives individual developers 180,000 code completions a month against Copilot's 2,000, and concluding that Google is the obvious pick for anyone cost-conscious. That was a fair summary once. It is not a fair summary now, and if you are pricing a rollout off it you will reach the wrong answer.
What actually changed on 18 June 2026
Google's own documentation is blunt about it: "Starting June 18, 2026, Gemini Code Assist IDE Extensions and Gemini CLI stopped serving requests for the Gemini Code Assist for individuals, Google AI Pro, and Google AI Ultra tiers. Affected users should migrate to Antigravity and Antigravity CLI."
Read that carefully, because the wording is stronger than a deprecation notice. The extensions stopped serving requests. Not throttled, not warned, not moved to a smaller allowance. If your developers were on the free individual tier, or on a personal Google AI Pro or Ultra subscription, the thing they were using does not work any more, and there is no downgrade path back into it.
So the free-tier column in every comparison table is now empty on the Google side. The paid editions, Standard and Enterprise, carry on unchanged, and they are what you are actually choosing between when you put Code Assist next to Copilot.
The paid entry seats, side by side
| Gemini Code Assist Standard | GitHub Copilot Business | |
|---|---|---|
| Entry price | $19 per user, per month, 12-month commitment | $19 per user, per month |
| How the vendor publishes it | Hourly: $0.026027397 per hour | Monthly, per seat |
| No-commitment price | $22.80 per user, per month | $19, no annual commitment required |
| Next tier up | Enterprise, $45 per user, per month | Copilot Enterprise, $39 per user, per month |
| Included AI allowance | No credit meter; daily request caps instead | 1,900 AI credits, worth $19.00 at $0.01 each |
| Are code completions metered? | Capped at 6,000 requests per user per day | No, completions are not billed in credits at all |
| Agent limits | 1,500 agent mode and CLI requests per user per day, combined | Agent work draws on the AI credit balance |
| Repository support | Any repo, in a supported IDE | GitHub |
| How you buy it | Gemini Admin console or sales, against a Google Cloud billing account | Directly on GitHub |
The first row is the one people find surprising. Two of the largest software companies in the world independently landed on the same number for their entry business seat, and it has been $19 on both sides for long enough that it is clearly a market price rather than a coincidence.
Why Google's price is written per hour
If you go looking for that $19 on Google's pricing page you will not find it. What is there is a table of license fees per hour, given to nine decimal places: $0.026027397 for Standard on a 12-month commitment, $0.031232877 on a monthly commitment, $0.061643836 and $0.073972603 for Enterprise on the same two terms.
Multiply the first by the 8,760 hours in a 365-day year and you get $228.00 exactly. The others give $273.60, $540.00 and $648.00. Four ugly numbers that all resolve to clean annual figures, which tells you Google set a round annual price and divided it by the hours in a year. Code Assist licenses are Google Cloud SKUs, and Google Cloud meters SKUs hourly.
There is a practical consequence nobody mentions. An hourly license on a monthly invoice means the bill tracks the calendar: 744 hours in a 31-day month, 720 in a 30-day month, 672 in February. The same Standard seat therefore invoices $19.36, $18.74 and $17.49 across those three months. It is small money per developer and it is a genuine nuisance when finance is matching invoices against a purchase order that says $19. Copilot does not have this problem, because GitHub bills a flat seat. We work through the full arithmetic, including the team-size tables, on our Gemini Code Assist pricing page.
Where the two actually diverge
Since the entry prices are identical, the decision has to be made on something else. Three things matter, roughly in this order.
Where your code is hosted. Copilot is a GitHub product. Its best features, the coding agent, the pull request summaries, the code review surface, assume your repositories are on GitHub. Code Assist works with whatever repository you have open in VS Code, IntelliJ or Cloud Shell Editor, so a team on Bitbucket, self-hosted GitLab or anything else is not shut out. For most organizations this single question settles it before price is even discussed.
Which cloud you are already in. Code Assist reaches into BigQuery, Cloud Run, Firebase, Colab Enterprise and Database Studio without any setup, and you buy it on a Google Cloud billing account you probably already have. If you are not a Google Cloud shop, none of that is worth anything to you, and the purchasing motion becomes an obstacle rather than a convenience. A small team with no Google Cloud footprint has real setup work to do before the first $19 is spent.
How the meter is shaped. This is the subtle one. Copilot Business gives you 1,900 AI credits worth a cent each, which comes to exactly $19.00 of included AI on a $19 seat, and crucially GitHub does not bill code completions against that balance at all. The meter only funds chat and agent work. Code Assist has no credit meter, but it does have hard daily caps: 6,000 code generation and completion requests, 960 chat requests, and 1,500 agent mode and Gemini CLI requests combined, all per user per day. Neither model is obviously better. One can produce an unexpected bill, the other can produce an unexpected stop.
Should you pay for the Enterprise tier on either?
Usually not, and for different reasons on each side.
Gemini Code Assist Enterprise is $45 against Standard's $19, so 2.37 times the price. Look at Google's own edition matrix and every feature a developer touches while writing code is on both: completion, generation, chat in the IDE, local codebase awareness, code transformation, agent mode, the Gemini CLI, and gen AI indemnification. What Enterprise adds is code customization against your private repositories, which is genuinely useful if your codebase has strong internal conventions, plus Gemini in Apigee, Gemini in Application Integration and Gemini Cloud Assist. Three of those four are separate Google Cloud products bundled into a developer seat. The daily agent cap rises from 1,500 to 2,000, a 33% increase for a 137% price increase.
Copilot Enterprise at $39 doubles the AI credit allowance to 3,900 and adds knowledge bases and organization-wide customization. Whether that is worth $20 a developer depends almost entirely on whether your teams will actually use the knowledge base feature, which in our experience is the thing most often bought and least often configured. We work through that decision properly in Copilot Business against Copilot Enterprise.
What the free tier ending should actually prompt
If you are reading this because your developers lost access in June and you now need to spend money where you previously spent none, it is worth pausing before defaulting to the nearest paid seat.
An in-editor assistant and an autonomous agent are different purchases that got bundled together, and the bundling hides which one you actually wanted. Autocomplete makes a developer faster at typing code they were already going to write, and it is worth paying per seat for, because every developer needs their own. An agent that takes a ticket, changes the code, runs the tests and opens a pull request is not per-developer work at all. It is throughput, and paying for it by headcount is the wrong unit.
That distinction shows up in the daily caps. Code Assist allows 1,500 agent requests per user per day, which sounds enormous until you notice it is shared with the CLI and that Google warns one prompt can produce several model requests. It is an assistant with an agent mode attached, priced as an assistant. If what your team is short of is hands rather than typing speed, the honest options are a dedicated agent or, for a genuinely under-resourced team, bringing in developers and autonomous agents on demand rather than buying another seat for everyone.
A short answer for each situation
Your code is on GitHub. Buy Copilot Business. Same price, the integrations are deeper on the surface you already use, and completions are not metered.
Your code is not on GitHub. Copilot is not really on the shortlist. Code Assist Standard at $19 works with whatever repository is open in your editor, and if you are already on Google Cloud the rest of the integrations come free.
You are on GitLab. Both of these are the wrong shape, and GitLab's own AI is now metered rather than a seat. GitLab Duo pricing covers what that costs per developer.
You need a fixed annual number. Copilot gives you one cleanly. Code Assist gives you $228 or $540 per seat per year, and no monthly figure that will match an invoice.
You wanted the work done, not the suggestion. Neither of these is really that product. Agentcode connects to a GitHub or GitLab repository on any hosting, takes a described task, plans it, edits the code, runs your existing test suite and opens a pull request for a human to review. It never merges on its own, and it costs a flat $29 a month billed yearly rather than $19 multiplied by everyone.
If you are still assembling the shortlist, every AI coding tool priced side by side has the rest of the field, and GitHub Copilot pricing breaks down the credit meter in detail.
Keep reading
- Jules vs Codex pricing: which async coding agent should you actually buy in 2026?
- Best AI coding CLI for an enterprise security review: Claude Code, Codex, Copilot and Cursor compared
- Cursor CLI vs Codex CLI: install, pricing, usage limits and which terminal agent to buy
- GitHub Copilot CLI vs Claude Code vs Codex CLI: what one terminal session costs on each