Claude Sonnet 5.5 Arrives in GitHub Copilot
Anthropic's midrange coding model reached every paid Copilot plan on September 28, promising the same work in fewer steps, tokens and tool calls.
In short
Claude Sonnet 5.5 has been available in GitHub Copilot since September 28, 2026, for the Pro, Pro+, Max, Business and Enterprise plans, billed at the provider's list price.
At a glance
- Live since September 28, 2026 in Copilot Pro, Pro+, Max, Business and Enterprise; the rollout is staged.
- Supported surfaces include VS Code, Visual Studio, JetBrains, Xcode, Eclipse, the Copilot CLI and the coding agent.
- List price: $2 per million input tokens, $10 per million output tokens, $0.20 per million cached tokens.
- Anthropic claims 30%+ more speed and up to 30% lower cost per task compared with Sonnet 5.
- Terminal-Bench 4.0: 70.6 percent, which Anthropic places above Opus 5.5 at 66.4 percent (xhigh).
Claude Sonnet 5.5 showed up in the GitHub Copilot model picker on September 28, 2026, for anyone on Pro, Pro+, Max, Business or Enterprise. GitHub says the rollout is gradual, so an account that does not offer the model yet is waiting rather than excluded. Anthropic aims it at well-scoped daily work: shipping a feature, closing a bug.
Where you can pick it
The surface list covers most of the places Copilot already lives: Visual Studio Code, Visual Studio, JetBrains IDEs, Xcode, Eclipse, the Copilot CLI, the Copilot coding agent, the Copilot app, github.com and GitHub Mobile on iOS and Android.
On Business and Enterprise, access runs through the model policy in Copilot settings. New models arrive enabled by default, so an administrator who wants tighter control has to opt out before the rollout reaches the organization rather than react afterward.
What it costs and what it claims
Anthropic's list price, as reported by heise online, is $2 per million input tokens and $10 per million output tokens, with cached reads at $0.20 per million tokens. The company says the model runs more than 30 percent faster than Sonnet 5 and can cut cost per task by up to 30 percent. Inside Copilot it is charged through usage-based billing at that list price.
On benchmarks, Anthropic reports 70.6 percent on Terminal-Bench 4.0, which it places ahead of Opus 5.5 at 66.4 percent on the xhigh setting. It also claims ten points more than Sonnet 5 on FrontierCode at high compute, for roughly one fifteenth of the cost per task. These are vendor figures, not independent evaluations.
GitHub's framing is more modest and more useful to a team lead: the same coding quality as Sonnet 5, reached with noticeably fewer steps, tokens and tool calls, and in less wall-clock time.
One rough edge
Developer Simon Willison reports a bug in the extended thinking settings. At the "max" effort level, one of his runs burned 128,000 tokens, cost $1.28, ran out of context and failed. The same task at "xhigh" finished in 41 seconds for 5.74 cents.
Still unanswered
GitHub's note does not publish a premium-request multiplier and points to its pricing documentation instead. None of the three sources checked here states a context window size, so this article does not quote one. Outside Copilot, the model is reachable through the Claude platform, Amazon Bedrock, Google Vertex AI and Microsoft Foundry.
FAQ
Which Copilot plans include Claude Sonnet 5.5?
Copilot Pro, Pro+, Max, Business and Enterprise. GitHub is rolling access out in stages, so the model does not appear in every account at the same moment.
How much does Claude Sonnet 5.5 cost?
Anthropic lists $2 per million input tokens, $10 per million output tokens and $0.20 per million cached tokens. In Copilot it is charged through usage-based billing at that list price.
Is Sonnet 5.5 better than Opus 5.5?
On Terminal-Bench 4.0 Anthropic reports 70.6 percent for Sonnet 5.5 against 66.4 percent for Opus 5.5 at xhigh. That is one vendor-run benchmark, not a general verdict.
