Anthropic launches Claude Sonnet 5 as a cheaper way to run agents

Read the original article →

What happened

Anthropic released Claude Sonnet 5 on June 30, calling it its most agentic mid-tier model to date. It became the default model for Free and Pro users and is available across the Max, Team and Enterprise plans.

Introductory pricing is set at $2 per million input tokens and $10 per million output tokens through August 31, rising to $3 and $15 afterward. The model scores 63.2% on an agentic coding benchmark against Opus 4.8's 69.2%, while slightly outperforming Opus 4.8 on a knowledge-work benchmark. Anthropic also highlighted lower rates of undesirable behavior, hallucination and sycophancy, along with better resistance to prompt injection.

Why it matters

A mid-tier model that lands close to the flagship on many tasks, at a fraction of the cost, changes the math for running agents at scale. Anthropic prices Sonnet 5 below Opus 4.8, GPT-5.5 and Gemini 3.1 Pro.

The safety improvements matter because more capable agents are being handed longer, less supervised tasks. Better prompt-injection resistance directly reduces one of the main risks in autonomous workflows.

MintedBrain take

For most day-to-day agent work, the practical question is where a cheaper model becomes good enough. Sonnet 5 looks like a sensible default; reserve Opus for the harder coding jobs where the benchmark gap actually shows up.

References

This article was originally published at Anthropic. For the full piece, read the original article.

Discussion

  • Loading…

← Back to News