GitHub Copilot vs Claude Code: Copilot bills Claude Opus 5.5 at Anthropic's own $4 and $20 per million tokens, so you may not need both

GitHub Copilot bills Claude Opus 5.5 at $4 input and $20 output per million tokens, as Anthropic lists (October 9, 2026). Plans set the bill.

Claude Enterprise adds $20 per user to usage. Copilot Business includes $19 of usage in its $19 price.

Table 1. Plans that list Claude Opus 5.5, by monthly price and included usage at list rates. Claude plans show the default model in Claude Code. Data collected 2026-10-09 (UTC).

Copilot Max returns $2.00 of usage per $1 of price, the highest in Table 1.

Copilot Business and Copilot Enterprise return $1.00.

Anthropic publishes no dollar value for any Claude plan. Copilot Pro is the only plan in the table without Claude Opus 5.5.

How the prices and limits were collected

Prices and limits come from GitHub's Copilot plans page and from Anthropic's own pricing and Claude Code pages. All were collected on October 9, 2026. LLM Waves Research did not run either tool. Every source address and retrieval date sits in the raw data file, and collection follows the LLM Waves methodology.

This comparison ranks on monthly cost for one user first and on where each tool runs Claude second.

The chart plots one user's monthly cost as Claude usage grows from $0 to $250.

Claude Enterprise sits $20 above Copilot Business at every usage level from $19 upward.

Copilot bills each Claude model at Anthropic's own list price

GitHub turns the token cost of each request into AI credits at $0.01 a credit. GitHub's models and pricing table lists Claude Opus 5.5 at $4 input, $0.20 cached input, $5 cache write, and $20 output per million tokens.

Anthropic's price page lists the same four rates for Claude Opus 5.5. It matches GitHub on all 11 Claude rows that appear on both pages, 44 figures in all.

That includes Claude Opus 5 at $5 input and $25 output, covered on the Claude Opus 5 model page.

Table 2. GitHub's rates for five Claude models, in dollars per million tokens, matched against Anthropic's price page. Data collected 2026-10-09 (UTC).

GitHub also lists Claude Sonnet 4 and Claude Opus 4.8 (fast mode) (preview).

Anthropic's page does not list them, so both sit outside the match count.

A Claude Opus 5.5 token costs the same inside Copilot as on the Claude API. The price gap comes from the plans.

Anthropic publishes no dollar value for Claude Pro, Max, or Team usage

Anthropic's plan page lists Claude Max at 5x or 20x the usage of Claude Pro. It lists a Claude Team Premium user at 5x a Standard user. It gives no dollar or token amount for any of them.

GitHub publishes credits for every Copilot plan, so each Copilot plan has a dollar value.

Base credits equal the plan price in dollars on all five Copilot plans. Pro at $10 buys 1,000 base credits, Pro+ at $39 buys 3,900, Max at $100 buys 10,000, Business at $19 buys 1,900, and Enterprise at $39 buys 3,900.

Flex credits sit on top and add $5 to Copilot Pro, $31 to Copilot Pro+, and $100 to Copilot Max. GitHub's billing page describes the flex allotment as a variable part of included usage.

The chart splits included usage per $1 of price into base credits and flex credits.

Without flex, every Copilot plan returns exactly $1.00 of usage per $1.

According to GitHub's changelog, Claude Opus 5.5 is on Copilot Pro+, Max, Business, and Enterprise. It is not on Copilot Pro.

Anthropic's model settings page states that Claude Code defaults to Claude Opus 5.5 on Pro, Max, Team, and Enterprise.

Claude Pro at $20 a month is the cheapest listed way to run Claude Opus 5.5 in Claude Code. Copilot Pro+ at $39 is the cheapest in Copilot.

A Claude Enterprise user costs $20 a month more than a Copilot Business user at the same usage

GitHub's billing documentation pools included credits across an organization. Copilot Business includes 1,900 credits, worth $19, for each user at $19.

Usage above the pool bills at $0.01 a credit. Additional usage is on by default. A user who consumes $200 of Claude usage costs $19 plus $181 of extra usage, or $200 in all.

Claude Enterprise costs $20 per user per month when you pay for the year, plus usage at API rates.

The same $200 of usage costs $220. The Copilot price is spent as usage, so the $20 gap holds at every usage level from $19 upward.

The chart multiplies the $20 gap by 12 months and by team size.

A 50-user team pays $1,000 more each month, or $12,000 more each year, on Claude Enterprise.

Anthropic's Claude Code cost guidance puts enterprise use at $150 to $250 per developer per month.

Cursor vs Claude Code covers that range. At those levels, 50 users cost $7,500 to $12,500 a month on Copilot Business and $8,500 to $13,500 on Claude Enterprise.

Copilot Enterprise at $39 includes $39 of usage and is designed for enterprises on GitHub Enterprise Cloud. This comparison did not collect that cloud price.

The same cost page describes Team and Enterprise usage as a seat allowance on 5 hour and weekly windows, while the price page lists Enterprise usage at API rates.

The $20 gap rests on the price page.

Claude runs inside Copilot on three surfaces

GitHub Copilot users can reach Claude in VS Code, on GitHub and in JetBrains IDEs.

VS Code bills the model you pick for the next turn

The Claude agent in VS Code runs on Anthropic's Claude Agent SDK. VS Code's agent documentation states that the selected model sets the provider and the billing method for the next turn.

A model routed through Copilot bills the Copilot subscription. A model on your own Anthropic key bills Anthropic.

GitHub's cloud agent bills AI credits and Actions minutes

GitHub's third party agents page states that the Claude coding agent is in public preview on all paid Copilot plans. It also states that each session uses GitHub Actions minutes and AI credits.

On October 9, 2026, GitHub's Claude agent page lists Auto, Claude Opus 4.7 and Claude Sonnet 4.6. The third party agents page lists Auto, Claude Sonnet 4.6, Claude Sonnet 5, Claude Opus 4.8, Claude Opus 4.8 (fast mode) (preview), Claude Opus 5 and Claude Fable 5.1.

Neither list names Claude Opus 5.5.

JetBrains runs the Claude Code command line tool as an agent

Microsoft's JetBrains plugin wiki lists Claude Code as an agent provider. It is switched on in settings and needs Claude Code version 2.1.114 or newer. The wiki page states nothing about billing or sign in.

One session on Claude Opus 5.5 costs 600 Copilot credits

Tokens set the bill on both sides, so one worked example carries the comparison. Take a session that sends 1,000,000 input tokens and 100,000 output tokens to Claude Opus 5.5.

Every request is under 100,000 tokens, with no caching. Input is 1 million tokens at $4 per million, so $4.00. Output is 0.1 million tokens at $20 per million, so $2.00.

The session costs $6.00, or 600 credits.

The chart prices the same session on five Claude models.

The same session on Claude Fable 5.1 costs 1,500 credits, a full Copilot Pro allowance, and Copilot Pro does not list that model.

Anthropic's cost page states that thinking tokens bill as output tokens. It also states that thinking cannot be turned off on Claude Opus 5.5, Claude Sonnet 5.5, Claude Haiku 5.5 or the Claude Fable models.

The example counts none, so a real session costs more.

Claude Opus 5 costs 750 credits for the same session at the $5 and $25 rates set out in Claude API pricing.

Claude Code stops at a window limit, and Copilot stops at zero credits

Anthropic's cost page states that subscription usage resets on a rolling 5-hour window. A separate weekly limit is shared across all models. After a model-specific message, such as a limit on Opus, a developer can keep working by switching to a model outside that family.

Claude Code version 2.1.234 or later can wait for the reset and continue the task.

Copilot Pro and Pro+ users on existing annual plans keep request based billing, covered in how premium requests ended for most plans.

On monthly individual plans, GitHub's billing page lists three exits from an exhausted allowance: upgrade for the price difference, pay for more usage under a budget, or wait for the monthly reset.

Business and Enterprise pools reset at 00:00 UTC on the first day of each month. Unused credits are forfeited. GitHub's organization billing page documents no automatic fallback to lower cost models when a budget runs out.

The chart counts how many of those 600 credit sessions each plan's monthly allowance covers.

Copilot Pro covers 2.5 sessions a month, and Copilot Max covers 33.3.

Which plan fits which reader

Solo developer paying personally. Situational: Claude Pro at $20 a month is the cheapest plan that defaults Claude Code to Claude Opus 5.5. Copilot Pro+ at $39 is the cheapest Copilot plan that lists it, with $70 of included usage.

Anthropic publishes no dollar value for Claude Pro, so cost per task has Insufficient data.

Engineering lead budgeting 5 to 50 users. Best value: Copilot Business. At equal usage it costs $20 a user less than Claude Enterprise, or $1,000 a month at 50 users.

Claude Team seats list no dollar value for usage, so any cost comparison there has Insufficient data.

Existing user of one tool. Situational: a Copilot Pro+ user already has Claude Opus 5.5 in VS Code and on GitHub from one credit pool. Adding Claude Code adds a second bill.

A Claude Code user who adds Copilot Pro pays $10 and gets $15 of usage, without Claude Opus 5.5.

Admin in a regulated company. Situational: GitHub's plans page states that Anthropic retains prompts and outputs by default on Claude Fable 5 and Claude Fable 5.1.

Zero retention is available by request through the end of 2026, and other Claude models run under zero retention.

Anthropic's own retention terms for Claude Code were not collected, so that side has Insufficient data.

What this comparison did not measure

This article did not run a latency test, a throughput test, or an output quality benchmark against any model discussed. It does not rank code quality.

It leaves out the GitHub Enterprise Cloud price, the usage allowance behind each Claude Team seat, and the token use of each tool's own agent on the same task.

Does GitHub Copilot use Claude models?

Yes. GitHub's plan table lists 11 Claude models. They include Claude Opus 5.5 on Copilot Pro+, Max, Business and Enterprise, and Claude Sonnet 5.5 and Claude Haiku 5.5 on Copilot Pro.

GitHub bills each Claude model at the rates Anthropic lists.

Which is cheaper, GitHub Copilot or Claude Code?

At equal usage, Copilot Business costs $20 a user less than Claude Enterprise. For one developer, Claude Pro at $20 a month costs $19 less than Copilot Pro+ at $39.

Anthropic publishes no dollar value for Claude Pro usage, so cost per task cannot be compared.

Can you use Claude Code with a GitHub Copilot subscription?

GitHub documents Claude inside Copilot as a Claude agent in VS Code, a Claude coding agent on GitHub, and a Claude Code setting in JetBrains.

Anthropic's VS Code and JetBrains pages list a paid Claude plan or a Claude Console account for sign-in.

No page collected here describes signing in to Claude Code with a Copilot subscription.

Is Claude better than GitHub Copilot for coding?

This comparison does not rank code quality. Both tools can run Claude Opus 5.5 at the same $4 and $20 rates, so the model name is shared and the surrounding agent differs.

That agent was not measured here, so the answer rests on price, plans and limits only.

One real task on each tool shows what a table cannot

GitHub's usage dashboard lists credits used and credits left. The /usage command in Claude Code lists plan usage.

Running the same task on both, in your own repository, gives a figure no table here can.