Publication date: 1 July 2026
Anthropic released Claude Sonnet 5 (model ID claude-sonnet-5) on 30 June 2026, positioning it as its most agentic Sonnet model to date and a lower-cost way to run autonomous agents. Claude Sonnet 5 ships with a 1 million token context window, scores 85.2% on SWE-Bench Verified, and launches at introductory pricing of $2 per million input tokens and $10 per million output tokens. It is now the default model for Free and Pro users on claude.ai.
What can Claude Sonnet 5 do?
Claude Sonnet 5 is built for agentic work rather than single-turn answers. Anthropic describes it as a model that plans a sequence of steps, calls tools, reads the result, and continues without a human prompting it at each stage. In practice that means driving browsers and terminals, running multi-step coding tasks, and completing long-running workflows autonomously.
The model has a 1 million token context window and a maximum output of 128,000 tokens, which can be raised to 300,000 tokens through the batch-API beta header output-300k-2026-03-24. Its training data has a knowledge cutoff of January 2026. The large context window lets Claude Sonnet 5 hold entire codebases, long documents, or extended tool-use histories in a single session, which matters for agents that need to track state across many steps.
Claude Sonnet 5 benchmarks and technical specs
Anthropic reported a broad set of benchmark results for Claude Sonnet 5. On coding, it scores 85.2% on SWE-Bench Verified, 63.2% on SWE-Bench Pro, 78.3% on SWE-Bench Multilingual, and 38.8% on FrontierCode v1. On agentic and tool-use tasks it reaches 80.4% on Terminal-Bench 2.1, 81.2% on OSWorld-Verified, and 84.7% on BrowseComp as a single agent, rising to 86.6% in a multi-agent setup.
On reasoning and knowledge benchmarks, Claude Sonnet 5 scores 57.4% on Humanity’s Last Exam with tools, 79.5% on USAMO 2026, and 1618 Elo on GDPval-AA v2. In the health domain it reaches 57.8% on HealthBench Professional. Taken together, the results place Claude Sonnet 5 near the top of the Sonnet line for coding and agentic tasks while keeping the price point below Anthropic’s flagship Opus models.
How does Claude Sonnet 5 compare to Claude Opus 4.7 and GPT-5.6?
Within Anthropic’s own line-up, Claude Sonnet 5 sits below Claude Opus 4.7, the flagship model released in April 2026 for complex reasoning and long-running agent workflows. Sonnet models are designed to deliver most of that agentic capability at a lower cost per token, which is the main argument Anthropic makes for Claude Sonnet 5: comparable agent performance for a fraction of the flagship price.
Against competitors, the closest recent release is OpenAI’s GPT-5.6 family (Sol, Terra and Luna), which entered limited preview on 26 June 2026 with prices ranging from $1/$6 to $5/$30 per million tokens. Where GPT-5.6 Sol is currently restricted to a small set of approved partners, Claude Sonnet 5 is generally available immediately across consumer and developer surfaces, which is the clearest practical difference for teams choosing between them today.
Claude Sonnet 5 availability and pricing
Claude Sonnet 5 is available now. It is the default model for Free and Pro users on claude.ai and is live in Claude Code, the Claude API, Cursor, VS Code, and GitHub Copilot. Developers can call it with the model ID claude-sonnet-5.
Introductory pricing is $2 per million input tokens and $10 per million output tokens through 31 August 2026, after which standard pricing of $3 per million input tokens and $15 per million output tokens applies. The introductory rate makes Claude Sonnet 5 one of the cheaper options for running high-volume agent workloads before the price increase takes effect.
Where to find Claude Sonnet 5
Claude Sonnet 5 is Anthropic’s agentic Sonnet model with a 1M-token context window, 85.2% on SWE-Bench Verified, and introductory pricing of $2/$10 per million tokens, released on 30 June 2026. Full details are on the official Anthropic Claude Sonnet page.
Full technical details and model weights are available on the Anthropic website.
