Adding GHCP vs Claude Code pricing comparison blog post#180
Open
dk-klein wants to merge 3 commits into
Open
Conversation
This post compares the value of GitHub Copilot and Claude Code based on their billing structures and token usage, providing a detailed analysis of what $200 buys in terms of output tokens and heavy tasks.
Revised text for clarity and consistency throughout the document, including updates on Claude's pricing and usage metrics.
Clarified the explanation of Anthropic's token limits and adjusted the conclusion regarding the comparison between Claude Code and GitHub Copilot.
TomProkop
reviewed
Jun 11, 2026
|
|
||
| On June 1, 2026, GitHub Copilot switched from flat pricing to **token based billing** ([GitHub blog](https://github.blog/news-insights/company-news/github-copilot-is-moving-to-usage-based-billing/)). For agentic work the change was large. Some developers posted screenshots of monthly bills going from **tens of dollars to hundreds**. Others said it was not as bad as it looked. | ||
|
|
||
| That leaves a concrete question for anyone doing heavy agentic work on their own $200 a month. How much real work does GitHub Copilot deliver for that money, and how does it compare to Claude Code? The two tools do not meter the same way. Copilot counts credits spent, Claude Code counts a rolling time window. Anthropic does not publish how big that window is, so there is nothing to look up, and that is why this comparison exists. The honest way to do it is to measure how many tokens the window actually holds, price the same work at the per token rates Copilot bills, and see what $200 buys on each side. Most public answers are estimates with no method behind them. This post measures it on a real account and shows every step. |
Member
There was a problem hiding this comment.
- sounds like AI slop
- we should be cautious with using "window" word because people will confuse it with "context window" . context window sizes are the same
- GitHub charges 1:1 Anthropic rates as described below. It is very important to introduce the reader to the fact that the models/context sizes are exactly the same. The only difference is that if you pay for subscription plan (not direct API calls), you might get more tokens for the same money depending on the pattern of use.
|
|
||
| That leaves a concrete question for anyone doing heavy agentic work on their own $200 a month. How much real work does GitHub Copilot deliver for that money, and how does it compare to Claude Code? The two tools do not meter the same way. Copilot counts credits spent, Claude Code counts a rolling time window. Anthropic does not publish how big that window is, so there is nothing to look up, and that is why this comparison exists. The honest way to do it is to measure how many tokens the window actually holds, price the same work at the per token rates Copilot bills, and see what $200 buys on each side. Most public answers are estimates with no method behind them. This post measures it on a real account and shows every step. | ||
|
|
||
| Anthropic also shipped a new model partway through this work, Claude Fable 5. A new model in the rotation can move usage limits, so the task was measured a second time after it landed, **still on Sonnet**, to confirm the window held. Both runs are in the results. |
Member
There was a problem hiding this comment.
How is this relevant? You can use the same models in both harnesses. Ofc we need to compare apples to apples.
|
|
||
| ## The claims in circulation | ||
|
|
||
| The same numbers come up again and again, but the sources behind them are either missing or have no clear methodology. |
Member
There was a problem hiding this comment.
linkedin bro AI tone of voice
| - Claude Max 20x is "worth about $5,000 of API compute, around 25 times the price." This traces to one developer's estimate, passed around pricing blogs with no measurement shown ([findskill](https://findskill.ai/blog/claude-code-subscription-pricing-guide/)). | ||
| - Max 20x gets "roughly 220,000 tokens per 5 hour window." This is stated as a flat figure with no method behind it ([faros](https://www.faros.ai/blog/claude-code-token-limits)). | ||
|
|
||
| The gap exists because **Anthropic does not publish a token number for the window** ([usage limits](https://support.claude.com/en/articles/11647753-how-do-usage-and-length-limits-work)), and it describes limits in relative terms rather than fixed token counts. The number is not unknowable though. Community tools like ccusage read Claude Code's own local logs and report the tokens and the API equivalent cost of each run. That is the method this post uses. |
Member
There was a problem hiding this comment.
which window? if context window, then it doesn't make sense. Anthropic does publish price of i/o tokens for direct api use. all other harnesesses than CC/CD/CW need to use api billing with Antropic models. in the past they allowed you to use the subscription but they started detecting it and blocking it (clawbot/openclaw drama)
|
|
||
| ## How the two meters work | ||
|
|
||
| Both tools charge by tokens. A token is a small chunk of text. Every request mixes a few kinds, and the comparison depends on which one each side counts: |
Member
There was a problem hiding this comment.
this contradicts statements above.
- CC subscription is flat rate + additional optional overage. the CC subscription does not have fixed number of tokens
- Anthropic models can be billed via API rates which GitHub Copilot subscriptions use
| Both tools charge by tokens. A token is a small chunk of text. Every request mixes a few kinds, and the comparison depends on which one each side counts: | ||
|
|
||
| - **input**, what you send to the model | ||
| - **output**, what the model writes back |
Member
There was a problem hiding this comment.
model doesnt write. it sends response. it is stateless with caching on client and server
| - **cache read**, context reused from earlier and billed much cheaper than input | ||
| - **cache write**, context stored so it can be reused, billed a little more than input | ||
|
|
||
| Copilot meters every token at the model's own rate, the same price Anthropic charges on its API, and converts the total to AI credits at **$0.01 each**. GitHub publishes the figures it uses ([GitHub docs](https://docs.github.com/en/copilot/reference/copilot-billing/models-and-pricing)): |
Member
There was a problem hiding this comment.
As does Anthropic when you exceed the limits on the subscription?
|
|
||
| Claude is different. The $200 Max 20x plan is **flat**, and usage is gated by a rolling five hour window ([higher limits](https://www.anthropic.com/news/higher-limits-spacex)). Anthropic gives no token number for that window, only that **Max 20x is twenty times the Pro plan** ([Max plan](https://support.claude.com/en/articles/11049741-what-is-the-max-plan)). | ||
|
|
||
| The five hour window is not the only ceiling. Anthropic also caps usage over **a rolling week**, again with no published token number. This post measures both. The results show the weekly cap sitting well above what a normal working day reaches, so the five hour window is the one that decides the figures below. |
Member
There was a problem hiding this comment.
Add disclaimer that it is a situation to a specific date and Anthropic keeps changing it?
|
|
||
| The five hour window is not the only ceiling. Anthropic also caps usage over **a rolling week**, again with no published token number. This post measures both. The results show the weekly cap sitting well above what a normal working day reaches, so the five hour window is the one that decides the figures below. | ||
|
|
||
| These limits used to bite harder. Through 2025 Claude capped usage aggressively and throttled during peak hours. In May 2026, after a compute deal with SpaceX, Anthropic **doubled the five hour limits and removed the peak hour throttling** for Pro and Max ([higher limits](https://www.anthropic.com/news/higher-limits-spacex)). So the window a 9 to 5 developer runs into today is more generous than the one the older blog numbers were written against. |
Member
There was a problem hiding this comment.
as commented above: "window" is problematic
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Measured comparison of what $200 a month buys on Copilot vs Claude Code after Copilot's switch to usage billing. Short version: about 20x more for a solo 9 to 5 dev, measured on a real account and fully sourced.