Anthropic released Claude Sonnet 5, a new AI model designed to handle complex tasks autonomously—making plans, using tools, and working without constant human input. The model performs nearly as well as Anthropic's larger, more expensive Opus 4.8 but costs significantly less than competing models from OpenAI and Google, signaling that the AI industry's next battleground is cost and reliability rather than raw capability alone.
Summaries like this, in your inbox every morning.
Sign up free →What happened
Anthropic released Claude Sonnet 5, a midsize model with agentic capabilities—the ability to plan, use tools like browsers and terminals, and run tasks autonomously. It launches as the default for free and Pro plans starting Tuesday, priced at $2 per million input tokens and $10 per million output tokens through August 31, then $3 and $10 respectively.
Why it matters
Agentic AI—systems that work without constant human direction—has become table stakes across foundation model companies. Sonnet 5 performs close to Anthropic's larger Opus 4.8 model but costs less than Opus 4.8, OpenAI's GPT-5.5, and Google's Gemini 3.1 Pro, shifting competition from capability to cost and reliability. Developers can now choose between Sonnet 5 for lower cost or Opus 4.8 for higher accuracy on complex judgment calls.
What to watch
On an agentic coding benchmark, Sonnet 5 scores 63.2%, versus Opus 4.8's 69.2% and its predecessor Sonnet 4.6's 58.1%. It also demonstrates lower rates of unsafe behaviors like deception and hallucination compared to Sonnet 4.6, though not yet at Opus 4.8's safety level. Testers report it completes multi-step workflows—like updating Salesforce records and sending announcements—end-to-end, tasks prior versions would have abandoned midway.
No comments yet. Be the first to share your thoughts!
Log in to join the discussion
Get curated AI news from 200+ sources delivered daily to your inbox. Free to use.
Get Started FreeFree · takes 30 seconds · unsubscribe anytime
1 minute a day. The AI essentials.
200+ sources · Email / LINE / Slack
