Three subscriptions, three billing systems, and a set of plan terms that decide the bill more often than the sticker price does. This comparison prices ChatGPT, GitHub Copilot and Claude from the vendors’ own pages, checked on 21 October 2026, and answers one question: what do you actually pay, and for which shape of work is each one worth it?
Three things in the current price lists decide most of the money.
First, which product “Copilot” means. GitHub Copilot and Microsoft 365 Copilot are different products with different prices and different licensing. Most 2026 comparisons quote Microsoft. This page prices GitHub Copilot, because that is the one developers buy, and it names Microsoft separately where the two collide.
Second, plan terms, not headline rates. A seat is not a person. ChatGPT Business requires at least two seats, Claude Team sells in pairs from two to 150, and Claude Enterprise carries a 20-seat minimum. Those minimums, not the per-seat rate, set your first invoice.
Third, three different meters for overage. ChatGPT caps usage and sells credits on top. Claude holds a flat price and throttles, with optional API credits. GitHub Copilot meters everything in AI credits at a cent each and bills a pooled overage. “Hidden costs” is not one thing.
Prices and plan terms as published on 21 October 2026. US dollars, before tax. A subscription seat does not include API tokens; the two are billed separately.
Which Copilot is in the table
GitHub Copilot is the coding product, sold per seat. Its current list is Free at $0, Pro at $10 a month, Pro+ at $39, Max at $100, Business at $19 per granted seat per month, and Enterprise at $39 per granted seat per month (GitHub Copilot plans, 21 October 2026). Copilot Free is limited to individual developers who do not already have Copilot through an organization or enterprise, so it is not a fallback for a company that has bought seats.
Microsoft 365 Copilot is a different product: an add-on for Office apps, listed at $30.00 per user per month paid yearly on Microsoft’s enterprise pricing page, and it requires a separate qualifying Microsoft 365 license (Microsoft, 21 October 2026). A comparison that quotes “Copilot at $21 a seat” or “Copilot at $30 plus your M365 licence” is not describing the product priced below.
What one person pays
| ChatGPT | Claude | GitHub Copilot | |
|---|---|---|---|
| Free | $0. Unlimited text chats with GPT-6; limits on uploads, images, voice, deep research, memory and Codex | $0, free for everyone | $0, limited access |
| Cheapest paid tier | Go, $8/month (US) | None — Pro is the first paid tier | Pro, $10/month |
| Main paid tier | Plus, $20/month | Pro, $20/month, or $200 billed up front ($16.67/month) | Pro+ $39/month; Max $100/month |
| Heavy-use tier | Pro, $100, $200 or $500 per month | Max, from $100/month, in 5x and 20x Pro usage tiers | Max, $100/month |
| Annual option | None on Plus; annual only on Business and Enterprise | Pro and Team: monthly or annual | One published per-seat monthly rate, no annual price published |
Sources: ChatGPT pricing, ChatGPT Business overview, ChatGPT Go announcement, Claude pricing, GitHub Copilot plans. All fetched 21 October 2026.
OpenAI’s pricing page states the structure in one line: monthly plans exist for Go, Plus and Business, and annual plans only for Business and Enterprise. GitHub publishes a single per-seat monthly rate for Business and Enterprise, with no annual price on the pages checked.
The only real consumer annual discount in this comparison is Anthropic’s. Claude Pro on monthly billing is $20, so twelve months is $240. Annual billing is $200 up front. That is $40 a year, or 16.7% of the $240 you would otherwise pay. ChatGPT has no annual plan on Plus, so there is no saving to compute there, and Copilot publishes no annual rate. An earlier version of this page claimed savings of 16% on ChatGPT and 20% on Claude; both were arithmetic errors, because twelve months of a $20 subscription is $240 and the $192 Claude figure was never a Claude price.
What a team actually pays
| ChatGPT Business | Claude Team | GitHub Copilot Business | |
|---|---|---|---|
| Standard seat | $20/seat/month billed annually, $25 monthly | $20/seat/month billed annually, $25 monthly | $19 per granted seat per month |
| Premium seat | $100 annually, $125 monthly, 5x the usage and no 5-hour limit | $100 annually, $125 monthly, 5x the usage | No premium seat tier published |
| Seat range | 2 to 200 | 2 to 150 | Not published |
| Minimum purchase | 2 seats in total: two standard, or one standard plus one premium | 2 seats | Not published |
| Usage that comes with it | Usage limits; Business and Enterprise can purchase workspace credits for more | Up to $500/month in pooled API credits | 1,900 AI credits per user, which is $19 at one cent per credit |
| API usage | Not included, billed separately | Not included; the $500 pooled credits are for the Claude Platform | Not included; Copilot meters its own usage in AI credits |
Sources as above, 21 October 2026.
Two seats, billed annually: ChatGPT Business is 2 x $20 = $40 a month, or $480 a year. Claude Team is 2 x $20 = $40 a month, or $480 a year. GitHub Copilot Business is 2 x $19 = $38 a month, or $456 a year. The two chat products are identical on price and identical on their two-seat minimum; Copilot is $2 a month below them, which is $24 a year, and $1 a seat a month, or 5% of the $20 annual rate.
The same two seats on monthly billing cost 2 x $25 = $50 a month, or $600 a year. Annual billing therefore saves $120 a year for two seats, which is 20% of the $600 monthly-billing baseline. That saving is available from OpenAI and Anthropic alike.
At twenty seats the spread widens and the minimums start to bite. ChatGPT Business is 20 x $20 = $400 a month, or $4,800 a year. Claude Team is also $400 a month, but it stops at 150 seats. Copilot Business is 20 x $19 = $380 a month, or $4,560 a year, and Copilot Enterprise is 20 x $39 = $780 a month, or $9,360 a year.
Enterprise is where the published prices stop. Claude Enterprise is $20 per seat per month billed annually with a 20-seat minimum, so the floor is 20 x $20 = $400 a month, or $4,800 a year, before any usage: Claude Enterprise bills usage at API rates on top of the seat. ChatGPT Enterprise has no published price at all — the pricing page says contact sales — so the only defensible statement is that its cost is unknown until a quote arrives. GitHub publishes no seat minimum for Copilot Enterprise, so no floor can be derived from the $39 rate.
If the seat price is only half the decision, the other half is what the work is. Our build-or-buy arithmetic for an AI agent sets out the second half, and what AI agents actually do covers the workloads the seat would be bought for.
Three different meters for going over
| ChatGPT | Claude | GitHub Copilot | |
|---|---|---|---|
| How heavy use is charged | Usage limits per plan; Business and Enterprise can buy workspace credits for more access | Flat subscription price, throttled by a rolling five-hour window and weekly caps; Pro and Max can add monthly API credits, Team includes up to $500 a month | AI credits: one credit is $0.01, each seat adds credits to a shared pool, usage beyond the pool is $0.01 a credit |
| What happens at the limit | You wait for the window to reset, unless you buy credits | You are throttled; optional usage credits keep you working at standard API rates | The allowance runs down, then you stop or you pay the $0.01 rate from a budget the admin sets |
| What the seat does not cover | API tokens | API tokens | API tokens, and a Claude subscription does not buy Claude inside Copilot |
Two structural facts do most of the work here. OpenAI’s own Business documentation says a ChatGPT Business subscription does not include API usage and that API usage is billed separately. And GitHub’s documentation says Copilot usage is measured in AI credits and that how many credits an interaction uses depends on the model you choose and the complexity of the task — while Anthropic sells its pooled credits for the Claude Platform. A Claude subscription and a Copilot seat are therefore two separate meters. Buying Claude Pro does not buy Claude inside Copilot, and buying a Copilot seat does not buy API tokens.
The published version of this page made this into a claim about percentages: that a typical business pays 42% more than advertised, and that overage fees are the biggest budget surprise for 68% of small businesses. Neither figure came with a source, and neither survives contact with the vendor terms above. The real mechanism is structural, and it is knowable in advance.
When the API is cheaper than the seat
Take one heavy month: 10 million input tokens and 2 million output tokens.
| Model | Input, 10M tokens | Output, 2M tokens | Total |
|---|---|---|---|
| Claude Haiku 5.5 | 10 x $0.10 = $1.00 | 2 x $0.50 = $1.00 | $2.00 |
| GPT-6 Luna | 10 x $0.10 = $1.00 | 2 x $0.50 = $1.00 | $2.00 |
| Claude Sonnet 5.5 | 10 x $2 = $20.00 | 2 x $10 = $20.00 | $40.00 |
| GPT-6.1 Sol | 10 x $2 = $20.00 | 2 x $10 = $20.00 | $40.00 |
| Claude Opus 5.5 | 10 x $4 = $40.00 | 2 x $20 = $40.00 | $80.00 |
| GPT-6 Astra | 10 x $10 = $100.00 | 2 x $50 = $100.00 | $200.00 |
| Claude Fable 5.1 | 10 x $10 = $100.00 | 2 x $50 = $100.00 | $200.00 |
Prices per 1M tokens from the Claude pricing page and OpenAI’s pricing cards, both fetched 21 October 2026. Haiku 5.5 rates above apply to prompts of 100K tokens or fewer; above that, input is $0.50 and output $2.50 per 1M tokens. Batch processing is half price, fast mode on Opus 5.5 is double, and US-only inference is 1.1x.
Read that against a $20 seat. At 10M in and 2M out, the same tokens cost $2 on Haiku 5.5 or GPT-6 Luna and $40 on Sonnet 5.5 or GPT-6.1 Sol. A $20 subscription is ten times a month of that volume on Haiku and half of it on Sonnet.
That gives a clean crossover. A $20 seat equals the API bill at roughly 5M input and 1M output tokens a month on Sonnet 5.5 or GPT-6.1 Sol (5 x $2 + 1 x $10 = $20), and at about 100M input and 20M output tokens on Haiku 5.5 or GPT-6 Luna (100 x $0.10 + 20 x $0.50 = $20). Below those volumes the seat is cheaper; above them the API is. Either way you need the API on top, because no seat includes API tokens.
The verdict, by who you are
| Situation | Cheapest on price alone | Monthly | Difference | What price does not decide |
|---|---|---|---|---|
| One person, occasional use | ChatGPT Free or Claude Free | $0 | — | Which models and how much usage the free tier carries |
| One person, everyday use | ChatGPT Go | $8 | Claude’s first paid tier is $20; Copilot Pro is $10 but is a coding tool | Whether Go’s limits fit your day |
| One person, heavy daily use | Claude Max from $100, or ChatGPT Pro at $100 | from $100 | Level | Models, usage limits, priority access — the price is identical |
| One person, willing to prepay a year | Claude Pro annual | $16.67 effective, $200 up front | $40 a year, 16.7% of $240; ChatGPT has no annual plan, Copilot no annual rate | Whether you need Pro rather than Plus |
| Two seats, code-heavy | GitHub Copilot Business | $38 | $2 a month below ChatGPT Business and Claude Team on annual billing; $12 below them on monthly billing | Whether a coding product is what the team needs |
| Two seats, general work | ChatGPT Business or Claude Team | $40 annual, $50 monthly | Identical seat prices and identical two-seat minimums | Which one your documents and code live in |
| Twenty seats | Copilot Business $380, or ChatGPT Business / Claude Team $400 | $380 to $400 | $20 a month between the cheapest and the chat plans; Claude Enterprise starts at the same $400 and adds API-rate usage on top | Whether you need seats, credits, or self-hosted hardware |
Each row follows from the tables above rather than from a general impression, and the $20 to $50 monthly band that most readers will pay is the same for two of the three products. The interesting decisions are not price decisions. They are product decisions: which workspace your work lives in, and whether your usage is heavy enough to meter separately.
If the real question is whether the agent work earns a seat at all, the build-or-buy arithmetic is the place to start, and if the answer is to run models on your own silicon, the hardware metrics that actually matter for AI agents cover the cost side of that decision. For the argument that a $20 subscription is not automatically the best value, see the $20 AI subscription myth.
What changed since this page was first published
This page was first published on 1 May 2026. Several figures on it did not match the vendors’ own pricing pages, and the savings percentages were arithmetically zero. Rather than patch the numbers and leave the structure alone, the version above was rebuilt from the vendor terms: seat minimums, billing cycles, what each vendor meters and what it does not.
One structural change is worth naming. Copilot used to be sold on a premium-request model with monthly request multipliers, and that model is now marked legacy in GitHub’s own documentation. It is metered in AI credits instead, and how Copilot’s 2026 pricing and model limits changed covers that history in detail.
Questions people actually ask
Is ChatGPT Plus cheaper than Claude Pro?
No, not by a cent. Both are $20 a month, and ChatGPT has no annual plan at that tier. Claude Pro’s annual option is $200 up front, which works out to $16.67 a month, a $40 annual saving. If you will prepay, Claude is the cheaper of the two; if you will not, they are the same price and the choice is about the product.
Is GitHub Copilot really cheaper than ChatGPT for a team?
On seat price, yes: $19 per granted seat against $20 standard seats, and $19 against $25 if you pay monthly. The gap is $1 a seat a month on annual billing, which is 5% of the $20 rate. A two-seat team pays $38 a month on Copilot Business and $40 on ChatGPT Business. That is the whole difference, and it is worth $24 a year for two people — real, but small.
What is the cheapest way to get a Claude subscription?
Claude Pro at $20 a month, or $200 billed up front. There is no paid Claude tier below Pro, so if $20 is too much, the free plan is the only option under it. Copilot Free is a separate matter: GitHub restricts it to individual developers who do not have Copilot through an organization or enterprise.
Do these plans include API usage?
No, none of them. OpenAI’s Business documentation states it directly, Anthropic sells API credits as a separate line and includes a pooled allowance on Team, and GitHub meters Copilot usage in its own AI credits. If your workload is metered by tokens, budget for it as a second line, and use the crossover table above to decide which side of it you are on.
Which is the best value for one person?
At the entry point, ChatGPT Free or Claude Free, both $0. For everyday use, ChatGPT Go at $8 is the cheapest paid tier of the three, and Claude Pro at $20 is the cheapest paid Claude plan. If you are doing heavy work every day, ChatGPT Pro and Claude Max both start at $100, so price stops deciding and the usage limits decide.
What about enterprise?
Only one of the three publishes a number that a buyer can act on. Claude Enterprise is $20 per seat per month billed annually with a 20-seat minimum, so $400 a month before usage, and usage is then billed at API rates. GitHub Copilot Enterprise is $39 per granted seat per month with no published minimum. ChatGPT Enterprise is contact-sales only, so any figure you see for it is a buyer report, not a price.
Independent technology writer focused on artificial intelligence, emerging technologies, and digital innovation. Covers AI applications in sports, productivity, and online business.
fstrtvlgwk
November 8, 2026 at 7:35 pm
yxmuhxqvjoqohllrlpemhhomxsfgoj