Claude Sonnet 5.5 in GitHub Copilot: Availability, Cost & SXF Analysis

Is Claude Sonnet 5.5 available in GitHub Copilot?

Yes. GitHub's September 28 release post calls Claude Sonnet 5.5 generally available. Its announced eligible subscriptions are Copilot Pro, Pro+, Max, Business and Enterprise. It is listed for Visual Studio Code, Visual Studio, Copilot CLI, Copilot coding agent, the Copilot app, GitHub.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse. The initial rollout was gradual; availability can still be affected by plan, app version and organizational settings. GitHub release note.

Where can you select it?

What actually changed in the September 28 GitHub release?

The change was not the invention of an entirely new Copilot product. It was the addition of Anthropic's Sonnet 5.5 as a selectable model inside existing GitHub Copilot workflows. GitHub said that in its early coding tests the model matched Sonnet 5 on its chosen tasks while needing fewer steps, tokens and tool calls, and finished noticeably faster. That is an internal GitHub observation; the post does not publish a full methodology, task corpus or reproducible per-task measurements.

For a developer this can matter in two distinct ways. First, a model that needs fewer tool calls can reduce interaction time and context overhead on routine changes. Second, the lower amount of work can reduce token spend, although GitHub's billing is driven by actual token consumption and rates. Neither observation establishes that Sonnet 5.5 is best for every monorepo, difficult debugging case or long-running agent task.

How to use Claude Sonnet 5.5 in Copilot

1. Check eligibility. Confirm a supported Copilot plan and sign in to the same account in GitHub.com or your editor.
2. Open the model selector. In a supported Copilot chat or agent workflow, open the model picker and select Claude Sonnet 5.5. Copilot CLI also documents claude-sonnet-5.5 as a model identifier and supports model selection via --model or /model; check the current CLI reference first.
3. Verify policy and scope. On Business and Enterprise, administrators can restrict models. If Sonnet is missing, verify the organization's model policies, the app version and supported surface before assuming the model was removed.

For full editor-specific setup, a model-picker walkthrough and troubleshooting, see SXF's dedicated Claude Sonnet in Copilot guide. Official references: Copilot CLI model commands and organization model policies.

How much does Claude Sonnet 5.5 cost in GitHub Copilot?

GitHub says Sonnet 5.5 uses provider list pricing under usage-based billing. Its current Copilot price table shows $2 per million input tokens, $0.10 per million cached-input tokens, $2.50 per million cache-write tokens and $10 per million output tokens for Sonnet 5.5. These are model unit rates, not a fixed per-prompt charge. GitHub converts measured usage into AI Credits, and subscription allowances, budget controls and the selected surface affect the final bill. GitHub's official pricing table.

Snapshot checked November 8, 2026. Haiku 5.5 figures in this table apply to GitHub's default context tier; long-context rates differ. Prices and supported models can change. Verify the linked live pricing table before making a purchase or budget decision.

Illustrative calculation, not a Copilot invoice: 100,000 uncached input tokens at $2 per million plus 20,000 output tokens at $10 per million equal $0.40 in model token charges before any cache writes, other model calls or platform-specific usage. A real agent task may generate many separate calls. To estimate costs accurately, look at the entire run, not just the user's visible prompt.

See the Claude Sonnet 5.5 model reference and SXF model pricing directory for broader model context. Do not conflate GitHub AI Credits with a separate Claude.ai subscription.

Is Sonnet 5.5 better than Sonnet 5 or Opus 5.5 for coding?

Anthropic's launch report says Sonnet 5.5 generates output over 30% faster than Sonnet 5 and uses fewer tokens for many tasks. It reports 70.6% on Terminal-Bench 4.0 against 10.3% for Sonnet 5 under Anthropic's stated evaluation setup. These results should be attributed to Anthropic; they are not an independent SXF replication. Different benchmark environments, tool policies and reasoning-effort settings can substantially change outcomes. Anthropic's launch evidence and caveats.

For complex, open-ended work requiring sustained judgment, Anthropic itself still describes Opus 5.5 as stronger. For small, repetitive or high-volume tasks, Haiku 5.5 may be the better cost candidate. A project's best default should follow its own error tolerance, review burden and budget—not the model family name. The site also tracks Opus 5.5 in Copilot and Haiku 5.5 in Copilot.

SXF analysis: the real advantage is fewer verified steps, not a larger benchmark number

SXF's view: Sonnet 5.5 is most interesting as a workflow-efficiency option for ordinary developer work. GitHub's reported reduction in steps and tokens matters only if it preserves correctness. A coding agent that completes a task quickly but introduces a regression creates downstream review and repair costs that a token benchmark does not measure.

A useful outcome measure is verified changes per dollar and per developer hour, not tokens per second alone. Consider review effort, escaped bugs, tool-call failures and time spent explaining or undoing a patch. That creates a different model ranking for a mature test-rich repository than for a fast-moving prototype with weak test coverage.

A practical evaluation protocol before your team switches

Rather than choosing one global default from a single demonstration, run a small, repeatable internal trial. SXF recommends the following decision protocol as an evaluation design, not as a measured result:

1. Define 20 representative tasks: eight scoped bug fixes, six feature additions and six larger debugging or refactoring tasks. Use identical repository snapshots and the same acceptance tests across candidate models.
2. Control the environment: match Copilot product surface, tool permissions, instructions, test harness and budget limits; record effort settings where exposed. Repeat at least some tasks to assess variability.
3. Grade independently: count fully passing changes, regressions, reviewer corrections and how often the model invents APIs or modifies unauthorized files.
4. Measure complete costs: log end-to-end elapsed time, model calls, AI Credits, tool failures and human review minutes. A failed first attempt is not free if a second model must repair it.
5. Choose by task group: use Sonnet for the class where it produces the best verified outcome; retain an escalation route for more complex work.

See SXF's AI Agent Reliability research for why single-run success and benchmark accuracy miss repeatability, robustness and real deployment risk. The GitHub Copilot alternatives guide helps compare product environments rather than models alone.

Security, access and organization policy: what model selection does not change

Sonnet 5.5 operates inside the selected Copilot feature. A chat suggestion, IDE agent, CLI action and cloud coding agent have different execution paths and permissions. Choosing Sonnet does not automatically allow filesystem writes, terminal commands, secret access or merges. Set repository and tool permissions separately, review proposed changes, and keep sensitive credentials out of prompts and output artifacts.

For Business and Enterprise, the organization can manage which models are available. GitHub documents separate Copilot policy and model controls, and enterprise rules can override organization choices. A “missing model” issue may therefore be governance rather than rollout. Official organization policy instructions. The Copilot memory guide covers context persistence and its practical boundaries.

Frequently asked questions about Sonnet in Copilot

What is Sonnet in Copilot?

Sonnet is an Anthropic Claude model tier. In GitHub Copilot, it is one of the selectable reasoning models; Copilot provides the interface, code context, tools and billing system.

Can I use Claude Sonnet 5.5 in GitHub Copilot for free?

The September 28 launch announcement lists Pro, Pro+, Max, Business and Enterprise, not Copilot Free, for explicit Sonnet 5.5 availability. Check current model support and billing before assuming access.

How do I enable Claude Sonnet 5.5 in VS Code?

Sign in with an eligible Copilot subscription, open Copilot Chat, select Claude Sonnet 5.5 from the model picker and verify that your account or organization permits that model.

Why is Sonnet 5.5 not showing in Copilot?

Check your subscription, signed-in account, model policy, editor extension version and whether the product surface supports the model. GitHub's release initially used a gradual rollout.

Is Claude Sonnet 5.5 cheaper than Opus 5.5 in Copilot?

On GitHub's published standard token rates, Sonnet 5.5 has lower input and output prices than Opus 5.5. Total task cost still depends on calls, tokens, cache use and retries; lower per-token prices do not guarantee a cheaper completed task.

Is Claude Sonnet 5.5 the same as Claude Code?

No. Sonnet is a model; Claude Code is Anthropic's separate coding product. Selecting Sonnet inside Copilot does not switch you into Claude Code.

Official sources, verification and related reading

This story is based on GitHub's public release note, its current product and billing documentation, and Anthropic's launch evidence. Product availability and pricing are point-in-time facts and may change; SXF's recommendations are explicitly marked as editorial analysis. Last verified: November 8, 2026.

1. GitHub Changelog: Sonnet 5.5 in GitHub Copilot — release date, eligible plans, supported surfaces and GitHub's early efficiency claim.
2. GitHub Docs: supported AI models — current product support.
3. GitHub Docs: models and pricing — AI Credits and per-million-token rates.
4. GitHub Docs: organization policies — administrative control over model access.
5. GitHub Docs: Copilot CLI command reference — model selection and identifiers.
6. Anthropic: Sonnet 5.5 announcement — model speed, benchmarks and limitations.
7. Anthropic system cards — model evaluation and safety documentation.

Go deeper: Sonnet in Copilot: setup, pricing and best uses · Sonnet 5.5 model details · Copilot alternatives · Agent reliability and evaluation.