Anthropic releases Claude Sonnet 5.5 as GitHub adds it to Copilot

Anthropic released Claude Sonnet 5.5 on September 28, positioning it as a faster and less expensive successor to Sonnet 5 for coding, office work and agentic tasks. The company says the model is available now across Claude, the Claude Platform and major cloud platforms, with GitHub also adding it to Copilot for paid individual, business and enterprise plans.

What changed

The announcement is mainly about efficiency rather than a new top tier model. Anthropic says Sonnet 5.5 runs more than 30 percent faster than Sonnet 5 and costs up to 30 percent less per task in its own testing, while keeping the same list prices of $2 per million input tokens, $10 per million output tokens and $0.20 per million cache read tokens. That matters because many teams are no longer asking only whether a model can solve a task. They are asking how often it can do that at production scale without turning every workflow into a surprise bill.

Anthropic’s benchmark claims are specific enough to make the release worth watching. The company reports a 70.6 percent score on Terminal-Bench 4.0, compared with 10.3 percent for Sonnet 5, and says the model is close to Opus 5.5 on several work and computer use tests. It also says Sonnet 5.5 is the first Sonnet model to launch with cyber safeguards and fallbacks similar to its most capable models because its cybersecurity ability has improved sharply.

Why developers will notice

GitHub’s rollout turns the model from a platform announcement into something many developers may encounter inside their daily editor. GitHub says Claude Sonnet 5.5 is generally available in Copilot and can be selected in Visual Studio Code, Visual Studio, Copilot CLI, the Copilot coding agent, github.com, GitHub Mobile, JetBrains IDEs, Xcode and Eclipse. The rollout is gradual, and business administrators can manage access through model policy settings.

For working teams, the practical question is not whether Sonnet 5.5 wins every benchmark. Anthropic itself says Opus 5.5 remains stronger for complex open ended work requiring sustained judgment. The useful dividing line is more operational: Sonnet 5.5 looks aimed at common coding tasks, bug fixes, document work and agent steps where speed and cost matter as much as maximum reasoning depth.

• Developers should compare it against their own repositories and review standards before switching defaults.
• Organizations using Copilot should check whether administrators have enabled new models automatically.
• Security teams should note the new cyber fallback behavior, especially where AI tools are used for vulnerability research or code review.

The most important test now is consistency in ordinary work. If Sonnet 5.5 really uses fewer tokens and fewer tool calls while staying reliable, it could shift more AI coding from occasional assisted work into always available background help. If the gains are uneven, teams may still reserve stronger models for architecture, audits and difficult debugging.

Sources

• www.anthropic.com
• github.blog

Cover photo by Anna Shvets on Pexels, used under the Pexels License.