Claude Code vs GitHub Copilot comes down to how much Claude you use. Copilot offers Claude Sonnet 5.5 from $10 a month and Opus 5.5 from $39, but since 1 June 2026 it charges every Claude chat and agent request at Anthropic's per-token list price, and a plan includes only $15 to $200 of that usage. Claude Code comes with Claude's $20, $100 and $200 plans, which limit usage by time window instead, so heavy Claude work costs far less there: a month of daily Opus 5.5 agent sessions comes to $182.24 on Copilot Max and $20 to $100 on a Claude plan.
Copilot still does things Claude Code doesn't: unlimited inline completions on paid plans, models from six makers in one subscription, and agents you start from a GitHub issue. Everything below was checked on 5 November 2026 against Claude Code 2.1.318, GitHub Copilot CLI 1.0.105 and VS Code 1.145. It covers GitHub Copilot, the coding product, not Microsoft 365 Copilot.
Claude Code vs GitHub Copilot at a glance
| GitHub Copilot | Claude Code | |
|---|---|---|
| What it is | Completions, chat and agents in IDEs, GitHub.com, a CLI and a desktop app | Anthropic's coding agent: terminal, IDEs, desktop app, browser |
| Models | Claude, GPT, Gemini, Grok, Kimi and Microsoft's MAI, plus Copilot Fusion routing | Anthropic's Claude: Opus 5.5 (default), Sonnet 5.5, Haiku 5.5, Fable 5.2 |
| Individual plans | Free, Pro $10, Pro+ $39, Copilot Max $100 | Pro $20 ($200 a year), Claude Max $100, Ultra $200 |
| How usage is limited | A monthly allowance priced per token ($15 to $200 of usage), then pay per token | A 5-hour and a weekly limit (Ultra: weekly only), shared with the Claude apps |
| Claude Opus 5.5 | Pro+, Copilot Max, Business, Enterprise | Every paid plan |
| Inline completions | Unlimited on paid plans, 2,000 a month on Free | Not a Claude Code feature |
| Cloud agent | Copilot in the cloud, from issues, pull requests or the Agents tab | Cloud sessions, and @claude through GitHub Actions |
| Team seat | Business $19, Enterprise $39, each including its own price in usage | Team Standard $25 ($20 yearly), Premium $125 ($100 yearly) |
| Free option | Copilot Free; Copilot Student for verified students | None; needs a paid plan or API key |
GitHub Copilot plans after premium requests
Copilot's plans changed on 1 June 2026, when GitHub replaced premium requests with usage-based billing. Under the old model, Pro included 300 premium requests a month and Pro+ 1,500, each prompt cost one request times a model multiplier, and extra requests cost $0.04. Most claude vs copilot comparisons written before June still count those requests.
Now every interaction is priced by its input, output and cached tokens at the model's rate. GitHub's unit is GitHub AI Credits, worth $0.01 each, so the table converts each allowance to dollars of list-price usage.
| Plan | Price | Included usage at list price | Claude models you can pick |
|---|---|---|---|
| Copilot Free | $0 | A limited allowance through auto model selection; 2,000 completions a month | None directly |
| Copilot Student | $0 for verified students | An allowance through auto model selection; unlimited completions | None directly |
| Pro | $10 a month | $15 ($10 base, $5 flex) | Haiku 4.5, Haiku 5.5, Sonnet 5, Sonnet 5.5 |
| Pro+ | $39 a month | $70 ($39 base, $31 flex) | Pro's, plus Opus 4.8, Opus 5, Opus 5.5, Fable 5, Fable 5.1 |
| Copilot Max | $100 a month | $200 ($100 base, $100 flex) | Same as Pro+ |
| Business | $19 a seat | $19 a seat, pooled across the organization | Same as Pro+ |
| Enterprise | $39 a seat | $39 a seat, pooled | Same as Pro+ |
How the allowance works, from GitHub's billing docs:
- The base part is used first, then the flex part, which GitHub's Max plan announcement says will vary over time. Nothing rolls over; allowances reset at 00:00 UTC on the 1st.
- Completions and next edit suggestions don't draw on the allowance on paid plans.
- When it runs out, you upgrade (paying the price difference), allow additional usage at the same per-token rates up to a spending limit, or wait. The old fallback to an included model is gone, and on Business and Enterprise additional usage is on by default.
- Paid plans get 10% off model costs with auto model selection, and 30% off with Copilot Fusion through 30 November 2026.
Annual Pro and Pro+ subscribers could keep premium requests until their term ends, with steeper multipliers (Claude Opus 4.8 costs 27 requests a prompt) and no new models, then drop to Copilot Free.
Claude models inside Copilot
Copilot charges Anthropic's list prices for Claude, token for token. These rates, from GitHub's models and pricing page, match Anthropic's API pricing:
| Model (per 1M tokens) | Input | Cache write | Cached input | Output |
|---|---|---|---|---|
| Claude Fable 5.1 | $10.00 | $12.50 | $0.25 | $50.00 |
| Claude Opus 5.5 | $4.00 | $5.00 | $0.20 | $20.00 |
| Claude Sonnet 5.5 | $2.00 | $2.50 | $0.10 | $10.00 |
| Claude Haiku 5.5, up to 100K | $0.10 | $0.125 | $0.01 | $0.50 |
So Copilot is a per-token route to Claude with a small allowance attached: Pro includes 1.5 times its price in usage, Pro+ 1.8 times, Copilot Max 2 times, and Business and Enterprise exactly their price. Claude Code justifies its own price once your Claude usage outgrows that allowance.
Claude runs in four places in Copilot:
- Chat and agent mode in VS Code and other IDEs, with the model picked per session.
- Copilot CLI, whose default model is
claude-sonnet-5.5, per its command reference. - The Anthropic Claude coding agent on GitHub.com (public preview), built on the Claude Agent SDK and billed to your plan plus Actions minutes. Its model list on 5 November named Opus 5, Sonnet 5 and Fable 5.1, but not Opus 5.5 or Sonnet 5.5.
- The Claude harness in VS Code, billed through Copilot or your own Anthropic API key.
The 1M-token context window for Opus 5.5 and Sonnet 5.5 is opt-in, in VS Code and Copilot CLI only, and uses more of the allowance.
What you can't do is bring a Claude plan into Copilot. Anthropic's Agent SDK docs bar third-party products from offering claude.ai login or plan rate limits unless Anthropic approves them. To use a Claude plan inside VS Code, install Anthropic's own Claude Code extension, which installs next to Copilot.
Claude Code plans and usage limits
Claude Code comes with every paid Claude plan on Anthropic's pricing page; the Claude Code pricing guide covers each in depth.
| Plan | Price | Limits |
|---|---|---|
| Free | $0 | No Claude Code |
| Pro | $20 a month, or $200 a year | A 5-hour limit and a weekly limit |
| Claude Max | $100 a month | 5x Pro, every 5 hours and every week |
| Ultra | $200 a month | 10x Pro every week; no 5-hour limit once the account is verified |
| Team | $25 Standard, $125 Premium a seat ($20 and $100 billed yearly) | Standard more than Pro; Premium 5x Standard |
| Enterprise | $20 a seat yearly, 20-seat minimum, plus all usage at API rates | Admin spend limits |
| API key | Per token, Opus 5.5 at $4 / $20 | Spend cap by API tier |
Max 5x and Max 20x became Max and Ultra on 30 October 2026, at the same prices. Chat and Claude Code share one pool; at a limit you wait, upgrade, or turn on pay-as-you-go overflow at API rates. Claude Code defaults to Opus 5.5 at medium effort, per its model docs.
Anthropic publishes multipliers, not token counts. SemiAnalysis's 5 October report found the same value per dollar on every Claude tier, and its team put a $200-a-month plan at about $12,000 of Opus 5.5 tokens at API prices on its workload. That's roughly 60 times the plan price if you use the whole allowance, against Copilot's 1.5 to 2 times.
Agent mode, coding agent and Claude Code: how the work differs
Copilot is built around your editor and GitHub: inline completions, Ask, Plan and Agent modes in VS Code chat, and Copilot CLI in the terminal, with -p for scripts and --cloud for a GitHub-hosted sandbox. Since 28 October, Copilot in the cloud starts sessions from an issue, a pull request or the Agents tab and continues them in the Copilot app, VS Code or the CLI. Each cloud task opens one pull request, and sessions on GitHub Actions stop at 59 minutes. Agent Merge fixes a pull request's failed checks and conflicts, and code review averages $0.25 to $10 a review by effort level, plus Actions minutes.
Claude Code starts from a task: it reads the repository, edits files and runs commands and tests, asking first as your permission mode requires, in the terminal, VS Code, JetBrains, the desktop app or the browser. Its overview covers CLAUDE.md, hooks, skills, subagents, MCP and claude -p. Cloud sessions count toward your plan's limits and can auto-fix a pull request's failing checks, and the GitHub Actions integration answers @claude in issues and pull requests, authenticated with your plan or an API key.
| Task | GitHub Copilot | Claude Code |
|---|---|---|
| Inline completions | Yes, unlimited on paid plans | No |
| Terminal agent | copilot | claude |
| Agent from a GitHub issue | Assign Copilot or the Claude agent | @claude through GitHub Actions |
| Pull request review | Copilot code review, billed per review | Code Review, or a review workflow in Actions |
| Model choice | Six makers, or Fusion picks per prompt | Claude models; others only through third-party endpoints |
| Project instructions | AGENTS.md or .github/copilot-instructions.md | CLAUDE.md; it also reads AGENTS.md |
| Per-run spend cap | --max-ai-credits in Copilot CLI | --max-budget-usd with an API key |
Both install in one line on macOS and Linux:
# Claude Code (native installer, updates itself)
curl -fsSL https://claude.ai/install.sh | bash
claude
# GitHub Copilot CLI (Node.js 22 or later); Opus needs Pro+ or higher
npm install -g @github/copilot
copilot --model claude-opus-5.5
How the two agents score
The independent Artificial Analysis Coding Agent Index (v2.0: Terminal-Bench 4.0, FrontierCode and SWE-Atlas-QnA) has run both. Costs are at API prices, not what a subscriber pays.
| Agent and model | Index score | Cost per task | Time per task |
|---|---|---|---|
| Claude Code, Opus 5.5 (max effort) | 61.7 | $11.42 | 61 minutes |
| GitHub Copilot, Copilot Fusion (max) | 59.9 | $5.59 | 36 minutes |
| Claude Code, Sonnet 5.5 (max effort) | 59.3 | $13.30 | 80 minutes |
Fusion, launched on 28 October, routes each prompt among Opus 5.5, Sonnet 5.5, GPT-6 Astra and GPT-6 Sol, and used about half Claude Code's tokens per task. It led Terminal-Bench 4.0 with 71; Claude Code led FrontierCode, 56 to 51. On Copilot, each $5.59 task draws from the allowance, so Copilot Max's $200 covers about 36 a month, or 51 with the November Fusion discount.
Monthly cost for the same usage
The three months below reuse the workloads from the Claude Code pricing guide: each agent step resends the conversation, 92% as cache reads and 8% as five-minute cache writes. Copilot charges Anthropic's rates, so the token bill is identical; what differs is how much each plan absorbs.
| Month | Workload | Claude tokens at list price, Opus 5.5 / Sonnet 5.5 |
|---|---|---|
| Light | 6 active days, 360 calls, 40,000-token context | $12.73 / $6.36 |
| Daily | 20 active days, 6,000 calls, 60,000-token context | $282.24 / $141.12 |
| Heavy | 22 active days, 26,400 calls, 90,000-token context, parallel agents | $1,757.18 / $878.59 |
Copilot costs the plan plus any usage past its allowance, on the lowest-cost plan that offers the model.
| Month | Copilot, Opus 5.5 | Copilot, Sonnet 5.5 | Claude Code |
|---|---|---|---|
| Light | $39.00 (Pro+) | $10.00 (Pro) | $12.73 with an API key, or Pro at $20 |
| Daily | $182.24 (Copilot Max) | $100.00 (Copilot Max) | Pro at $20, or Claude Max at $100 once the 5-hour limit gets in the way |
| Heavy | $1,657.18 (Copilot Max) | $778.59 (Copilot Max) | Claude Max at $100 or Ultra at $200 |

The break-even points, in Claude tokens at list price per month:
- Up to $25, Copilot Pro at $10 costs less than any Claude plan for Sonnet 5.5 agent work (under $10, an Anthropic API key costs less still). At Anthropic's average of $13 per developer per active day, $25 is about two active days.
- Past $25, Copilot Pro costs more than Claude Pro's $20. For Opus, Copilot's entry point is Pro+ at $39, already above Claude Pro.
- Past $131, Copilot Pro+ costs more than Claude Max. Past $200, Copilot Max does, and past $300 it costs more than Ultra.
The claude code vs copilot math assumes both agents use the same tokens for the same work, and they don't. One Copilot CLI user estimated Claude models take about 20% more turns in Copilot's harness. Fusion goes the other way: scaled by Artificial Analysis's cost per task ($5.59 against $11.42), the daily Opus month becomes about $138 of usage, which is $100 on Copilot Max, $107.15 on Pro+, or $65.71 on Pro+ while the 30% Fusion discount lasts. That puts Fusion in Claude Max's range, though Fusion picks the models for you and still costs more than Claude Pro's $20.
A team of ten
For ten developers on the daily workload, Copilot Business and Enterprise both come to the token bill, because each seat includes exactly its price in usage: $1,411.20 on Sonnet 5.5, $2,822.40 on Opus 5.5. A Claude Team plan costs $250 a month with Standard seats ($200 billed yearly) or $1,250 with Premium seats, if the seat limits hold the work. Claude Enterprise works like Copilot: $400 for the 20-seat minimum plus every token at API rates.
GitHub Copilot vs Claude Code: who should pick which

Pick GitHub Copilot if:
- You mainly want completions and chat, with agent tasks a few times a month. Pro at $10 covers about $15 of Sonnet 5.5 work.
- You want GPT-6.1, Gemini, Grok and Claude in one picker, or Fusion choosing for you.
- Your company already pays for Business or Enterprise and sets policies and spending limits there.
- You want agents that start from GitHub issues, next to Agent Merge and code review.
Pick Claude Code if:
- You run agents most working days, especially on Opus 5.5. A $20 to $200 plan holds months that cost $182 to $1,657 on Copilot.
- You want Claude Code's own tooling: hooks, skills, subagents, cloud sessions and the harness Anthropic builds for its models.
- You also use Claude for chat, which shares the same plan.
You can also run both: Copilot Pro for completions and Claude Pro or Claude Max for agent work, $30 or $110 a month, with one AGENTS.md that both agents read. For other pairings, see Cursor vs Claude Code, Codex vs Claude Code and Claude Code alternatives, which covers agents on other models.
What developers report
- GitHub's Chief Product Officer, Mario Rodriguez, on the billing change: "Agentic usage is becoming the default" (GitHub blog, 27 Apr 2026, link).
- "2 points below Opus 5.5 in Claude Code at half the Cost per Task" (Artificial Analysis on Copilot Fusion, 28 Oct 2026, link).
- A Hacker News commenter with Copilot CLI at work and a personal Anthropic subscription said Copilot CLI copies most new Claude Code features within weeks, but guessed Claude models need about 20% more turns in its harness (27 Apr 2026, link).
- "Copilot has improved immensely in 2026." (a Hacker News commenter comparing it with Claude Code, 28 Apr 2026, link)
- An annual Pro+ subscriber on Hacker News used Copilot mostly as a fallback when a Claude Max plan hit its limits, and described Opus's jump from 3x to 27x on annual plans as rough (28 Apr 2026, link).
- A Copilot CLI user reported the monthly allowance climbing from about 97.8% to 100% with no further prompts (GitHub issue #4308, 30 Jul 2026, link).
Where Aight fits
Claude Code can't run on Aight: it speaks Anthropic's Messages API, and Aight documents an OpenAI-compatible Chat Completions API. Copilot CLI's bring-your-own-key file does accept OpenAI-compatible providers, and sends those requests to the provider, which bills them.
Aight lists Claude Opus 5.5 and Sonnet 5.5 at Anthropic's list prices on its models page. A $100 recharge costs $103 and adds $110 of balance, about $0.94 per $1 of list-price usage, where Copilot's additional usage costs the full list price. Each key can have a monthly cap (403 key_cap_reached) and a list of allowed models. Copilot CLI needs tool calling and streaming, while Aight's API reference documents only model, messages and stream, so test a tool-using session first. Configured as an OpenAI-compatible provider in ~/.copilot/providers.json:
{
"providers": [
{ "name": "aight", "type": "openai", "baseUrl": "https://api.getaight.ai/v1", "apiKey": "sk-aight-..." }
],
"models": [
{ "id": "claude-sonnet-5-5", "provider": "aight" }
]
}
Then start it with copilot --model aight/claude-sonnet-5-5. For daily agent work, though, a Claude plan beats every per-token route: the daily Opus month costs $264.28 through Aight against $100 for Claude Max.
FAQ
Can GitHub Copilot use Claude Opus 5.5?
Yes, on Pro+, Copilot Max, Business and Enterprise. Pro has Sonnet 5.5 and Haiku 5.5 but no Opus, and Free and Student reach models only through auto selection. Opus 5.5 costs $4 / $20 per million input / output tokens in Copilot, the same as Anthropic's API.
Can a Claude Pro or Max plan be used inside GitHub Copilot?
No. Anthropic doesn't let third-party products offer claude.ai login or plan limits without its approval, so Claude inside Copilot bills through Copilot or an Anthropic API key. To use a Claude plan in VS Code, install Anthropic's Claude Code extension.
Does GitHub Copilot cost less than Claude Code?
For light use, yes: Copilot Pro at $10 includes $15 of usage at list price, about one active day of agent work. Past about $25 a month of Claude tokens, Claude Pro at $20 costs less, and a month of daily Opus 5.5 sessions costs $182.24 on Copilot Max against $20 to $100 on a Claude plan.
What happened to Copilot premium requests?
Monthly plans, Business and Enterprise moved from premium requests to per-token billing on 1 June 2026. Annual Pro and Pro+ subscribers keep premium requests until their term ends, at higher multipliers and without new models, then drop to Copilot Free unless they switch to a monthly plan.
Does Claude Code have code completions like Copilot?
No. Claude Code's documentation describes an agent that edits files and runs commands, with no inline completion feature. If you want completions, Copilot Pro adds them for $10, unlimited and outside the allowance.
Sources
- GitHub Docs: Plans for GitHub Copilot, Usage-based billing for individuals, Usage-based billing for organizations and enterprises, Models and pricing, Supported AI models, What's changing with GitHub Enterprise and Copilot billing
- GitHub Docs, legacy billing: What changed with Copilot billing, Requests in GitHub Copilot, Model multipliers for annual plans
- GitHub Docs, agents and CLI: Third-party coding agents, Copilot on GitHub.com, Copilot CLI, Installing Copilot CLI, CLI command reference, Session limits in Copilot CLI, Adding LLM models to Copilot CLI, Bring your own key, Copilot Fusion, Optimizing AI usage, Cloud sandbox billing
- GitHub blog: GitHub Copilot is moving to usage-based billing (27 Apr 2026) and Flex allotments and a new Max plan (12 May 2026); Copilot plans page; @github/copilot on npm; issue #4308
- Visual Studio Code: Choose an agent harness
- Anthropic: Plans and pricing, API pricing, Claude on X, 30 Oct 2026
- Claude Code docs: Overview, Manage costs, Model configuration, Cloud sessions, GitHub Actions, Commands, CLI reference, Agent SDK overview; @anthropic-ai/claude-code on npm
- Artificial Analysis: Coding Agent Index and GitHub Copilot Fusion analysis (28 Oct 2026)
- SemiAnalysis: Anthropic Subscriptions Offer 5x+ More Value Than OpenAI and post on X, 7 Oct 2026
- Hacker News: GitHub Copilot is moving to usage-based billing
- GitHub on YouTube: Everything announced at GitHub Universe 2026
- Aight: Models and pricing, API reference, Refund policy
*Checked on 5 November 2026. Updated 17 November 2026: Fable 5.2 is Claude Code's fable model; Claude Code 2.1.326 changes nothing else compared here.*