On 28 October 2026 GitHub made Copilot Fusion generally available. Instead of choosing a model, you choose an effort level and Copilot decides which models handle each task. GitHub's announcement also quotes a benchmark that puts Fusion Max close to Claude Code at about half the cost per task. This guide separates what is official from what is one source's measurement, and lays out what each option costs today, using GitHub's and Anthropic's own pages read on 28 October 2026.
What Copilot Fusion is and how it routes tasks
GitHub's changelog describes Fusion as a system you pick in the model picker in place of a specific model. For each prompt it decides which models to use and how they work together, and you only set the effort: Fusion Standard prioritises versatile models for everyday coding, Fusion High prioritises capable models for harder tasks, and Fusion Max prioritises the most capable models for the most demanding work.
Behind that sit three workflows. Single: one model solves the task. Cascade: an efficient model drafts, a quality check accepts it or escalates to a stronger model. Critique: one model drafts, another reviews, and the first revises. Fusion only uses models enabled in your plan and your organisation's policies. The Copilot plans page lists models from Anthropic, OpenAI, Google, xAI, Microsoft and others, but GitHub does not publish which ones Fusion picks for a given task.
Fusion works in VS Code, the GitHub Copilot app, Copilot CLI and JetBrains IDEs. GitHub's documentation table shows Fusion Standard on every plan, including Free and Student, while Fusion High and Fusion Max need Pro+, Max, Business or Enterprise.
Cost per task vs quality: what the benchmark says
The headline figure comes from the Artificial Analysis Coding Agent Index, as reported in GitHub's changelog of 28 October 2026: Copilot Fusion Max scored 60, "within two points" of Claude Code with Opus 5.5 Max. GitHub says it got there at about half the average API cost per task, with about half as many tokens and about 40% less time per task.
Treat this as one source's measurement on one index. GitHub's post gives the score and relative figures, but not a dollar cost per task, the task set or how the Claude Code run was configured, so we do not quote a dollar number. "Half the API cost" compares costs at API rates. Neither a Copilot subscription nor a Claude subscription bills you per task that way, so your real bill depends on your plan and how much you use it. Results on your own repositories can differ.
Pricing: Copilot plans vs Claude Code access
Copilot Fusion has no separate fee. GitHub's documentation says you are billed for each model it uses, at that model's standard rate, out of your GitHub AI Credits. Through 30 November 2026, users on paid Copilot plans get a 30% discount on model costs when they use Fusion. Claude Code, for its part, is included in every paid Claude plan and shares one usage pool with Claude chat; the alternative is pay-as-you-go API billing. We keep these prices, with the date each was read, in our AI tool price index.
| Option | Monthly price | What you get for coding | Limits and billing |
|---|---|---|---|
| Copilot Free | $0 | Fusion Standard, agent mode, Copilot CLI, 2,000 inline suggestions a month | Limited chat and agent usage |
| Copilot Pro | $10 | $15 of AI Credits ($10 base + $5 flex), Fusion Standard, coding agent, code review | No Fusion High or Max; extra credits can be bought |
| Copilot Pro+ | $39 | $70 of AI Credits ($39 base + $31 flex), all Fusion effort levels, delegation to Claude and Codex agents (preview) | Credit use varies by model |
| Copilot Max | $100 | $200 of AI Credits ($100 base + $100 flex), all Fusion effort levels | Flex allotments may change over time |
| Claude Pro | $20, or $17 a month billed yearly ($200 up front) | Claude Code plus Claude on web, desktop and mobile | One usage pool, 5-hour session limits and weekly limits |
| Claude Max | From $100 | Claude Code with 5x or 20x Pro usage, $100 to $200 a month in API credits | Billed monthly |
| Claude API | Per token | Opus 5.5: $4 input, $20 output per million tokens; Sonnet 5.5: $2 and $10 | You pay for every token you use |
Sources: github.com/features/copilot/plans, GitHub Docs "About GitHub Copilot Fusion" and claude.com/pricing, all read on 28 October 2026, USD before tax. Copilot Business, Enterprise and Student prices and the separate Claude Max 20x price were not shown on the pages we read, so they are left out.
Which one to pick: four scenarios
You already pay for Copilot Pro and mostly do everyday edits, tests and small fixes. Try Fusion Standard first. It is on your plan, it draws on the same AI Credits, and the 30% discount until 30 November makes this the cheapest time to see how far $15 of credits goes.
You run long, hard agent tasks and want the setup GitHub benchmarked. Fusion Max needs Copilot Pro+ at $39 or Max at $100. Compare that with Claude Max from $100, which is built around Claude models only. If you mostly want Claude anyway, Pro+ can also hand tasks to Claude's agent (in preview).
You work in the terminal and also use Claude for chat, documents or research. Claude Pro at $20 a month, or $17 billed yearly, covers Claude Code and the rest of Claude from one pool. The trade-off is limits instead of credits: when you hit the 5-hour or weekly limit you wait, upgrade, or turn on usage credits at API rates.
You automate coding in CI or scripts. Pay-as-you-go API pricing is the easiest to predict per task, because you can count tokens. At Opus 5.5 rates, a task that uses 200,000 input tokens and 20,000 output tokens costs about $1.20 (our calculation: 0.2 x $4 + 0.02 x $20). Sonnet 5.5 would halve that.
How to pay less
Use the lowest effort that does the job. Because Fusion bills each model at its own rate, Standard for routine work and Max only for hard problems is the simplest saving.
Measure before you commit. Run a week of real tasks on Fusion during the discount window, note the credits used per task, then compare with what the same work costs you on Claude.
Pay yearly if you keep Claude Pro: $200 up front instead of $240 over twelve monthly payments saves $40 (our calculation). Our AI subscription savings calculator shows what yearly billing saves across your other tools. On the API, prompt caching helps: cached input reads on Opus 5.5 cost $0.20 per million tokens instead of $4.
FAQ
Is Copilot Fusion free?
Fusion is available on every Copilot plan, including Free, where you get the Standard effort level. There is no separate Fusion fee: on paid plans each model it uses is billed at its standard rate from your AI Credits (GitHub Docs, read 28 Oct 2026).
Is Copilot Fusion better than Claude Code?
On the Artificial Analysis Coding Agent Index, as reported by GitHub, Fusion Max scored 60, two points below Claude Code with Opus 5.5 Max, at about half the API cost per task. That is one benchmark from one source, so test both on your own code.
Does Claude Pro include Claude Code?
Yes. Claude's pricing page says Claude Code is included in all paid plans and shares the same usage limits as the rest of the plan. Pro is $20 a month, or $17 a month billed yearly (read 28 Oct 2026).
How long does the 30% Copilot Fusion discount last?
GitHub's documentation says users on paid Copilot plans get a 30% discount on model costs when using Fusion through 30 November 2026.
Sources (read 28 Oct 2026)
- GitHub Changelog: Route and orchestrate models with GitHub Copilot Fusion (28 Oct 2026)
- GitHub Docs: About GitHub Copilot Fusion (read 28 Oct 2026)
- GitHub Copilot plans and pricing (read 28 Oct 2026)
- Claude plans and pricing, including API rates (read 28 Oct 2026)
- Kuponaut AI tool price index
- Kuponaut AI subscription savings calculator
Originally published on Kuponaut, where the prices are kept up to date.
Top comments (0)