<MCP.Playground/>
online

Claude Sonnet 5

Runs MCP toolsReasoning

Anthropic’s mid-tier model, released 30 June 2026 and the default on claude.ai’s Free and Pro plans. Anthropic calls it the most agentic Sonnet yet, so it is the Claude model most MCP servers meet first.

Vendor
Anthropic
Released
30 June 2026
Context
1M tokens
Input
Text, image, file

Paste a server URL, pick a model, and watch it call your tools in a real conversation. No install.

No server of your own? Leave it blank and use one of the public mock servers.

MCP Evals

Can Claude Sonnet 5 actually complete tasks with your server?

One chat shows the tool calls work. An eval writes real tasks from your tool schemas, has Claude Sonnet 5 drive each one, and reports where it got the wrong answer even though every call succeeded. Pick Claude Sonnet 5 as the driver.

Run evals →See a sample

What Claude Sonnet 5 is

Claude Sonnet 5 succeeds Sonnet 4.6. Anthropic describes it as able to make plans, use tools like browsers and terminals, and run autonomously, with performance close to Opus 4.8. It has a 1M-token window, takes text, image and file input, and supports selectable effort levels up to extra high. On Anthropic’s API it is claude-sonnet-5.

What Anthropic says

  • OSWorld-Verified: 78.5%.
  • Humanity’s Last Exam: 46.8% with tools, 34.6% without.
  • Performance close to Claude Opus 4.8 and a substantial improvement over Sonnet 4.6, by Anthropic’s account.

These are the vendor’s own claims, not measurements of ours. Run the model against your own server to find out whether they hold for your tools.

How it reaches MCP

  • Natively. Anthropic’s Messages API can connect to remote MCP servers itself. On the API the model id is claude-sonnet-5.
  • Through claude.ai, where it is the default model on the Free and Pro plans, and through Claude Code and Claude Desktop
  • MCP Playground’s Agent Studio: paste your server URL above and run Sonnet 5 against it in the browser, with no Anthropic key

What to watch for

  • Its Humanity’s Last Exam score rises 12 points when it can use tools. Tool descriptions carry a lot of weight with this model, so a vague description costs more than it would with a model that ignores tools.
  • As the claude.ai default, it is the model most of your users will connect with. A server that only passes on Opus or Fable will fail for them first.
  • OpenRouter lists neither temperature nor seed for it. Expect tool calls to vary between runs, and repeat a prompt before you trust one result.

Run the complex-schema mock on Sonnet 5 and Opus 5.5 with the same prompt. process_order has the deepest nesting of the four tools, which makes it the likeliest place for the smaller model to get an argument wrong. All the mock servers →

Available here

ModelModel ID
Claude Sonnet 5anthropic/claude-sonnet-5

Go deeper

Frequently asked questions

Does Claude Sonnet 5 support MCP?
Yes, natively. Anthropic’s API can connect to remote MCP servers directly, and claude.ai, Claude Code and Claude Desktop are all MCP clients. You can also run it against any server here without an Anthropic key.
Claude Sonnet 5 or Opus 5.5 for MCP tools?
Start with Sonnet 5. Anthropic puts it close to Opus 4.8, and for well-described tools that is usually enough. Run the same prompt on both against your server here. If the calls match, Sonnet is the one to ship.
What changed from Claude Sonnet 4.6?
Anthropic calls Sonnet 5 its most agentic Sonnet, able to plan and use browsers and terminals autonomously, with a substantial improvement over Sonnet 4.6. Both are available here if you want to compare them on your tools.
Is Claude Sonnet 5 free?
On claude.ai it is the default model on the Free plan. Here you can run it against any MCP server with no Anthropic key.

Sources

Claude Sonnet 5 for MCP: Specs & Tool Calling — Test Free | MCP Playground