GitHub Copilot Pro vs Pro+: Complete Comparison and Report on Price, AI Credits, Premium Models, and Agent Features

GitHub Copilot Pro costs $10 a month and GitHub Copilot Pro+ costs $39 a month. Both are individual plans that GitHub moved to usage-based billing on June 1, 2026, when every Copilot plan started drawing down GitHub AI Credits instead of counting premium requests. Since then the two plans differ in three measurable ways: the monthly credit allowance, the list of models a subscriber can pick by hand, and a short list of agent features that GitHub reserves for Pro+ and the $100 Max plan above it.
The verdict is clear for most developers. Pro+ includes 7,000 AI credits a month against 1,500 on Pro, 4.7 times the usage for 3.9 times the price, and it is the cheaper of the two as soon as a month of work needs more than 4,400 credits, because extra credits on Pro cost $0.01 each. Pro+ is also the only one of the two plans with Claude Opus 5.5, Claude Fable 5.2, GPT-6.1 Sol and GPT-6 Astra in the model picker, together with Fusion High, Fusion Max and third-party coding agent delegation. Pro remains the better buy for developers who mainly want unlimited code completions and next edit suggestions, which neither plan bills in credits, and who use chat or agent mode a few times a day with mid-priced models such as Claude Sonnet 5.5, GPT-6 Luna or Gemini 4 Hydrogen.
The decision therefore turns on two questions a reader can answer from their own usage page: how many credits a normal month consumes, and whether the work needs a frontier model that Pro does not list. Prices, allowances and model availability in this comparison were verified on GitHub's official plan pages, documentation and changelog on November 16, 2026.
··········
GITHUB COPILOT PRO AND PRO+ PRICES, CREDITS AND BILLING RULES.
GitHub's plan documentation lists Copilot Pro at $10 per calendar month, Copilot Pro+ at $39 and Copilot Max at $100. The pages verified for this comparison list no annual price for new subscribers. Each plan's allowance has two parts. The base allotment is fixed at the subscription price, so a $10 plan carries 1,000 base credits at the published rate of 1 AI credit = $0.01. The flex allotment is extra usage that GitHub adds on top and says "will vary over time" as model prices and efficiency change. GitHub announced the flex allotments and the Max plan on May 12, 2026, effective June 1.
........
Item | Copilot Pro | Copilot Pro+ |
Monthly price | $10 | $39 |
Base AI credits | 1,000 ($10) | 3,900 ($39) |
Flex AI credits | 500 ($5) | 3,100 ($31) |
Total included credits | 1,500 ($15) | 7,000 ($70) |
Included credits per dollar | 150 | 179 |
Extra usage | $0.01 per credit, by budget | $0.01 per credit, by budget |
Code completions and next edit suggestions | Unlimited, not billed | Unlimited, not billed |
Free for verified teachers and popular open-source maintainers | Yes | Not stated |
........
Credits reset at 00:00 UTC on the first day of each calendar month, whatever the subscription date, and GitHub's documentation states that included credits do not carry over between months. Base credits are consumed first and the flex allotment applies automatically after them, in the IDE, on github.com and in the Copilot CLI. When both are gone, a subscriber can upgrade, wait for the reset, or set a dollar budget for additional usage. GitHub warns that individual plans "may have additional credit limits based on usage patterns, billing history, and verification status", so extra spending on a personal account is not unlimited by default.
Plan changes follow the rules on GitHub's licensing page. An upgrade takes effect immediately and costs the new plan's price minus what was already paid for the current cycle, and credits already used that month count toward the larger allowance. A downgrade takes effect at the start of the next billing cycle and is generally not prorated.
GitHub also applies two discounts that reduce credit consumption on both plans. Paid plans get 10% off model costs when auto model selection picks the model in Chat or cloud agent sessions, and Copilot Fusion carries a 30% discount on model costs through November 30, 2026.
··········
HOW GITHUB COPILOT PRO AND PRO+ REACHED THE CURRENT LINEUP.
The two plans looked different in early 2026. On April 20, 2026, GitHub paused new sign-ups for Pro, Pro+ and Student, removed Opus models from Pro, and described Pro+ as offering "more than 5X the limits of Pro". Existing subscribers could cancel for a refund of the remaining subscription time until May 20.
The usage-based model arrived on June 1, 2026. From that date all Copilot plans bill on AI credits, Copilot Max became available as an upgrade for existing Student, Pro and Pro+ subscribers, and Copilot code review started consuming GitHub Actions minutes in addition to credits. On June 17, GitHub announced that sign-ups for Student, Pro, Pro+ and Max were reopening gradually over about two weeks, which is why both plans can be bought today.
........
Date | Change affecting Pro and Pro+ |
April 20, 2026 | New Pro and Pro+ sign-ups paused; Opus models removed from Pro |
May 12, 2026 | Flex allotments and the Copilot Max plan announced |
June 1, 2026 | Usage-based billing on AI credits for all plans; code review uses Actions minutes |
June 17, 2026 | Individual plan sign-ups start reopening over about two weeks |
September 22, 2026 | Claude Opus 5.5 added for Pro+, Max, Business and Enterprise |
October 28, 2026 | Copilot Fusion generally available; cloud sandboxes billed from November |
November 9, 2026 | Claude Fable 5.2 added for Pro+, Max, Business and Enterprise |
November 13, 2026 | Gemini 4 Hydrogen added for Pro and above; six models set to retire December 14 |
........
The sequence explains the present split. GitHub kept Pro as the entry plan with the full set of mid-priced models and moved every frontier model added since April onto Pro+ and higher.
··········
MODELS IN COPILOT PRO VERSUS COPILOT PRO+.
The "Supported AI models per Copilot plan" table in GitHub's documentation is the clearest source on what each plan can select. Every model available on Pro is also available on Pro+. The difference sits entirely at the top of the lineup.
........
Model | Copilot Pro | Copilot Pro+ |
Claude Opus 5.5 | Not included | Included |
Claude Opus 5 and Opus 4.8 (including Opus 4.8 fast mode preview) | Not included | Included |
Claude Fable 5.2 | Not included | Included |
Claude Fable 5 and Fable 5.1 | Not included | Included |
GPT-6 Astra | Not included | Included |
GPT-6.1 Sol and GPT-6 Sol | Not included | Included |
GPT-5.6 Sol and GPT-5.5 | Not included | Included |
Gemini 4 Argon | Not included | Included |
Claude Sonnet 5.5, Sonnet 5, Haiku 5.5 | Included | Included |
GPT-6 Luna, GPT-5.6 Luna, GPT-5.6 Terra | Included | Included |
Gemini 4 Hydrogen, Gemini 3.8 Flash, Gemini 3.7 Flash | Included | Included |
Grok 4.7, Kimi K3, MAI-Code-1.1-Flash | Included | Included |
GPT-5.3-Codex, GPT-5.4, GPT-5.4 mini, GPT-5 mini | Included | Included |
........
Claude Opus 5.5 reached Copilot on September 22, 2026, and the changelog entry names Pro+, Max, Business and Enterprise as the plans that receive it. Claude Fable 5.2 followed on November 9 with the same plan list and with data retention by default, because Anthropic keeps prompts and outputs to run safety classifiers. Gemini 4 Hydrogen, added on November 13, goes the other way and is available from Pro upward.
Two footnotes affect annual subscribers. Claude Sonnet 4.6 was retired on September 1, 2026, but remains available to individual subscribers on annual Pro and Pro+ plans, and Claude Haiku 4.5 will stay available to individual annual subscribers after its December 14 retirement. GPT-5.4 nano appears only in the Codex Visual Studio Code extension, which requires Pro+.
Free and Student users can reach models only through auto model selection. Pro and Pro+ subscribers can pick a model by hand, which is where the table above decides what each plan can do.
··········
GITHUB COPILOT MODELS RETIRING ON DECEMBER 14 FOR BOTH PLANS.
GitHub announced on November 13 that six models leave all Copilot experiences on December 14, 2026, a change covered in more detail in our report on the December 14 Copilot model retirements. The list touches both plans, and two of the retiring models are among the Pro+ exclusives.
........
Retiring model | Plans that list it today | Suggested alternative |
Claude Fable 5 | Pro+ and above | Claude Fable 5.2 |
Gemini 4 Argon | Pro+ and above | Gemini 4 Carbon |
Claude Haiku 4.5 | Pro and Pro+ (annual plans keep it) | Claude Haiku 5.5 |
Gemini 3.8 Flash | Pro and Pro+ | Gemini 4 Hydrogen |
Grok 4.6 | Pro and Pro+ | Grok 4.7 |
Kimi K3 | Pro and Pro+ | To be announced |
........
After December 14, Pro subscribers keep Gemini 4 Hydrogen as the replacement for Gemini 3.8 Flash and Grok 4.7 as the replacement for Grok 4.6, both already listed on Pro. The plan table verified for this comparison does not list Gemini 4 Carbon, so the documentation does not yet say which plans will receive the alternative GitHub suggests for Gemini 4 Argon. Our article on Gemini 4 Hydrogen in GitHub Copilot covers the pricing of the model that replaces Gemini 3.8 Flash on both plans.
··········
AGENT FEATURES, FUSION MODES AND TOOLS IN PRO AND PRO+.
Most Copilot features are shared. GitHub's plan comparison marks the CLI, the Copilot app, editor and IDE support, agent mode, MCP server integration, custom instructions and agents, Copilot Spaces and the public code filter as included on every individual plan. Pro adds to Free the cloud agent, Copilot code review, app modernization and issue-to-pull-request assignment, and Pro+ keeps all of them.
........
Feature | Copilot Pro | Copilot Pro+ |
Agent mode, MCP, custom instructions and agents | Included | Included |
Copilot cloud agent and cloud sandboxes | Included | Included |
Copilot code review | Included | Included |
Issue-to-pull-request assignment, app modernization | Included | Included |
Copilot Fusion Standard | Included | Included |
Copilot Fusion High and Fusion Max | Not included | Included |
Third-party coding agent delegation (preview) | Not included in the table | Included |
Codex Visual Studio Code extension with GPT-5.4 nano | Not included | Included |
Purchase of additional credits | Included | Included |
Pooled usage, SAML SSO, IP indemnity, usage metrics | Not included | Not included |
........
Copilot Fusion became generally available on October 28, 2026. It picks and orchestrates models instead of asking the user to choose one, with three modes. Fusion Standard "prioritizes versatile models for everyday coding", Fusion High prioritizes capable models for harder tasks and deeper reasoning, and Fusion Max prioritizes the most capable models for the most demanding work. Fusion bills each model it calls at that model's standard rate, and among individual plans GitHub limits the High and Max modes to Pro+ and Max. Our coverage of Copilot Fusion's general availability describes the routing in detail.
GitHub's plan page contains two inconsistencies that buyers should know about. The Pro card mentions access to third-party agents (Claude Code and Codex), while the comparison table marks third-party coding agent delegation as not included on Pro. The Pro+ card lists audit logs, which the comparison table does not include for any individual plan. Until GitHub aligns the two, the table is the more conservative reading for Pro.
Cloud sandboxes, the hosted environments behind the cloud agent, are available on paid Copilot plans only, and their billing began in November 2026 through GitHub's shared, dollar-based Actions compute entitlement. The changelog does not publish a separate sandbox allowance for Pro or Pro+.
··········
[[ADPLUS_300x250_1]]
INDEPENDENT EVALUATIONS OF THE MODELS ONLY PRO+ CAN SELECT.
The practical value of Pro+'s exclusive models can be measured with independent results. Artificial Analysis publishes an Intelligence Index (version 5.0.1) and agentic coding benchmarks for the main models on each side of the plan line. The figures below come from its comparison pages as read on November 16, 2026, at maximum reasoning effort unless stated, and were measured through the providers' APIs, not inside Copilot.
........
Artificial Analysis measure | Claude Opus 5.5 (Pro+) | Claude Sonnet 5.5 (Pro) | GPT-6.1 Sol (Pro+) | GPT-6 Luna (Pro) |
Intelligence Index v5.0.1, max effort | 59 | 55 | 55 | 28 |
Intelligence Index, high effort | 52 | 42 | 50 | 21 |
Terminal-Bench 4.0, max effort | 59.6% | 63.6% | 56.1% | 12.6% |
FrontierCode-AA 2.0, max effort | 46.4% | 41.7% | 40.6% | 25.2% |
API list price, input/output per 1M tokens | $4 / $20 | $2 / $10 | $2 / $10 | $0.10 / $0.50 |
........
The Claude pair shows a narrower gap than the price suggests. At maximum effort, Opus 5.5 leads Sonnet 5.5 by four points on the Intelligence Index and by 4.7 points on FrontierCode-AA 2.0, while Sonnet 5.5 scores higher on Terminal-Bench 4.0. The gap widens at lower effort. At high effort, Artificial Analysis records 56.6% for Opus 5.5 and 43.9% for Sonnet 5.5 on Terminal-Bench 4.0. The OpenAI pair is far apart at every setting, with GPT-6.1 Sol scoring 55 against 28 for GPT-6 Luna on the index and 56.1% against 12.6% on Terminal-Bench 4.0.
For a Pro subscriber, Claude Sonnet 5.5 is therefore the strongest everyday choice the plan allows, and at maximum effort it outscored Opus 5.5 on Terminal-Bench 4.0 in these tests. The models that justify Pro+ on quality grounds are Opus 5.5 for harder reasoning at moderate effort and GPT-6.1 Sol for developers who prefer OpenAI's flagship over GPT-6 Luna.
··········
WHAT A MONTH OF COPILOT WORK COSTS IN AI CREDITS.
GitHub prices every model in Copilot at provider list rates per million tokens and converts the result to credits at $0.01 each. A worked example makes the allowances concrete. The assumptions are one agent-mode task of 150,000 input tokens, of which 120,000 are read from the prompt cache and 30,000 are fresh, plus 8,000 output tokens. Fresh input is billed at the standard input rate and cache-write surcharges are left out, so tasks on models with a cache-write price would cost somewhat more. Prices are those in GitHub's "Models and pricing" table for prompts under the long-context threshold.
........
Model (plan) | Cost per task | Credits per task | Tasks within Pro's 1,500 credits | Tasks within Pro+'s 7,000 credits |
GPT-6 Luna (Pro and Pro+) | $0.0082 | 0.82 | 1,829 | 8,537 |
Gemini 3.8 Flash (Pro and Pro+) | $0.0615 | 6.15 | 244 | 1,138 |
Claude Sonnet 5.5 (Pro and Pro+) | $0.152 | 15.2 | 99 | 461 |
Grok 4.7 (Pro and Pro+) | $0.168 | 16.8 | 89 | 417 |
GPT-6.1 Sol (Pro+ only) | $0.152 | 15.2 | Not available | 461 |
Claude Opus 5.5 (Pro+ only) | $0.304 | 30.4 | Not available | 230 |
........
Two monthly profiles built on the same assumptions and 20 working days show where each plan fits. A developer who runs three Claude Sonnet 5.5 agent tasks and ten GPT-6 Luna chats a day uses about 1,076 credits a month, inside Pro's 1,500. A developer who runs fifteen Sonnet 5.5 tasks a day uses about 4,560 credits, which costs $40.60 on Pro with extra credits and $39 on Pro+. A developer who splits the day between six Claude Opus 5.5 tasks and six Sonnet 5.5 tasks uses about 5,472 credits, which only Pro+ or Max can serve because Pro does not list Opus 5.5.
Code review consumes credits at a different scale. GitHub estimates an average of $0.25 per review in Lite mode, $1.50 in Balanced mode and $10 in Max mode, excluding Actions minutes, with ranges of $0.05 to $1, $0.25 to $5 and $1.50 to $25. At those averages, Pro's 1,500 credits cover about 10 Balanced reviews or one and a half Max reviews a month, while Pro+'s 7,000 cover about 46 Balanced reviews or seven Max reviews.
··········
BREAK-EVEN POINTS BETWEEN COPILOT PRO, PRO+ AND MAX.
Because extra credits cost the same $0.01 on every individual plan, the cheapest plan for a given volume can be computed directly. Pro costs $10 plus $0.01 for every credit above 1,500, so it reaches Pro+'s $39 price at 4,400 credits. Pro+ costs $39 plus $0.01 for every credit above 7,000, so it reaches Max's $100 price at 13,100 credits.
........
Monthly credits used | Cost on Pro | Cost on Pro+ | Cost on Max | Cheapest plan |
1,500 | $10.00 | $39.00 | $100.00 | Pro |
3,000 | $25.00 | $39.00 | $100.00 | Pro |
4,400 | $39.00 | $39.00 | $100.00 | Pro and Pro+ equal |
7,000 | $65.00 | $39.00 | $100.00 | Pro+ |
10,000 | $95.00 | $69.00 | $100.00 | Pro+ |
13,100 | $126.00 | $100.00 | $100.00 | Pro+ and Max equal |
20,000 | $195.00 | $169.00 | $100.00 | Max |
........
The table assumes that a personal account can buy the extra credits it needs, which GitHub does not guarantee for every user given the additional credit limits on individual plans. It also assumes today's flex allotments. If GitHub lowers the flex portion, as its May 12 announcement allows, both break-even points move down, because the base credits stay fixed at the subscription price.
GitHub's own plan cards express the same ladder in ratios. The Pro+ card promises "4x+" the included usage of Pro, and the actual figures give 4.7 times. The Max card promises "2.9x+" the usage of Pro+, which matches 20,000 credits against 7,000.
··········
WHAT GITHUB HAS NOT PUBLISHED ABOUT PRO AND PRO+.
Several details that affect the choice are missing from GitHub's pages as of November 16, 2026. The documentation gives no credit amount for the Free and Student allowances, no annual price for new Pro or Pro+ subscribers, and no per-plan limits on Fusion High or Fusion Max beyond the credit allowance itself. The additional credit limits on individual plans are described only in general terms, without amounts.
The cloud sandbox allowance for Pro and Pro+ is not stated in the October 28 changelog, which refers readers to the general Actions entitlement. GitHub also has not said which plans will receive Gemini 4 Carbon, the alternative it suggests for Gemini 4 Argon after December 14, and the alternative for Kimi K3 is still listed as to be announced. The plan page's Pro card and comparison table disagree on third-party agent delegation, and the Pro+ card's audit logs entry is absent from the table.
··········
COPILOT PRO OR PRO+: WHICH PLAN FITS EACH TYPE OF DEVELOPER.
For readers comparing Copilot with Cursor, Claude Code and Codex before choosing a plan, our guide to AI coding agent pricing sets the Copilot plans against those tools. Within Copilot, the facts above point to a short set of choices.
........
Developer profile | Recommended plan | Reason |
Mostly inline completions, occasional chat | Pro | Completions and next edit suggestions are unlimited and not billed |
Daily agent mode with Claude Sonnet 5.5 or Gemini 4 Hydrogen, under about 4,400 credits | Pro | Cheaper than Pro+ below the break-even point |
Heavy agent mode above about 4,400 credits a month | Pro+ | 7,000 included credits cost less than Pro plus extra credits |
Needs Claude Opus 5.5, Fable 5.2, GPT-6.1 Sol or GPT-6 Astra | Pro+ | These models are not listed on Pro |
Wants Fusion High or Fusion Max routing | Pro+ | Pro includes Fusion Standard only |
Frequent Copilot code review in Balanced or Max mode | Pro+ | About 46 Balanced reviews a month against about 10 on Pro |
Consistently above 13,100 credits a month | Max | 20,000 included credits for $100 |
Verified teacher or popular open-source maintainer | Pro (free) | GitHub offers Pro at no cost to these users |
........
Pro at $10 is the right plan for a developer whose month stays below about 4,400 credits and whose work runs well on Claude Sonnet 5.5, which outscored Opus 5.5 on Terminal-Bench 4.0 at maximum effort in Artificial Analysis testing. Pro+ at $39 is the right plan once usage passes that line or once a project calls for Claude Opus 5.5, Claude Fable 5.2 or GPT-6.1 Sol, and its 7,000 credits also make it the cheaper way to run regular Balanced or Max code reviews. A subscriber unsure where they stand can start on Pro and upgrade mid-month, since GitHub charges only the price difference and counts the credits already used toward the Pro+ allowance.
··········
FOLLOW US FOR MORE.
·····
DATA STUDIOS
·····
datastudios.org




