Anthropic just launched Claude Sonnet 5, and the new model nearly matches Opus performance at a fraction of the price

Anthropic has released Claude Sonnet 5, a new model the company says narrows the performance gap with its more expensive Opus line. As reported by MacRumors, the model is described internally as the most agentic version of Sonnet that Anthropic has released to date. It is now the default model for both Free and Pro users on claude.ai.

According to Anthropic, Sonnet 5 is built to operate with greater autonomy, planning its own steps and using external tools such as terminals and browsers to complete tasks. The company says the model can finish complex jobs that its predecessor, Sonnet 4.6, was unable to handle, and that it now checks its own output without additional prompting. Anthropic has also reported improved resistance to malicious requests, along with lower rates of hallucination and sycophancy compared to earlier models.

In terms of raw capability, Anthropic positions Sonnet 5’s performance as close to that of Opus 4.8 across reasoning, coding, knowledge work, and tool use, according to the company’s official documentation. Opus models still retain an edge on certain high-level agentic tasks, but the proximity in performance comes at a notably lower price point.

The introductory pricing window will not last forever

Through August 31, 2026, Claude Sonnet 5 is priced at $2 per million input tokens and $10 per million output tokens. After that date, pricing shifts to $3 per million input tokens and $15 per million output tokens. The model is available across the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry. The release comes as Anthropic pursues other major moves in its business, including a confidential IPO filing submitted to the SEC earlier this year.

On the technical side, Sonnet 5 features a one million token context window and a maximum output of 128,000 tokens. It supports adaptive thinking, which runs at all times, and the model ID is a pinned snapshot rather than an evergreen pointer. On both the Claude API and Claude Code, the effort parameter defaults to high, though developers can adjust it manually.

Anthropic’s broader model lineup includes Claude Fable 5, which the company describes as its most capable widely released model for demanding reasoning and long-horizon agentic work. Claude Mythos 5 remains available only in limited form through Project Glasswing, primarily for research and cybersecurity workflows. Access to those top-tier models has drawn government attention in recent weeks, following an export control directive affecting Anthropic that limited access to its most advanced systems.

Claude Sonnet 5 is available now through claude.ai, the Claude API, Amazon Bedrock, Google Cloud, and Microsoft Foundry.