What's new

Yesterday (June 30, 2026) Anthropic released Claude Sonnet 5 — in its own words, its most agentic Sonnet to date. And in Ayeto you can assign it to any assistant right away, with no waiting and no migration.

Sonnet 5 succeeds Sonnet 4.6 and targets one specific thing: narrowing the gap between the Sonnet class and the pricier Opus class. Anthropic says its performance approaches that of Opus 4.8 — at a significantly lower price.

Why it matters

Sonnet-class models were where the agent era began for a lot of developers. Sonnet 5 keeps that going and, according to Anthropic, handles tasks where previous Sonnets tapped out halfway:

Agentic work. The model plans, uses tools like a browser and a terminal, and runs autonomously at a level that until recently only larger, pricier models could reach.
Software engineering. Early-access partners describe how Sonnet 5 writes a reproducing test on its own, fixes the bug, and verifies the fix — in a single pass, with no explicit instruction.
It checks its own work. Testers note that the model verifies its own output without being asked to.

That's a perfect fit for Ayeto: an assistant with tools (Python, REST API, Web Scraper, Microsoft 365…) gets a model that carries multi-step tasks through to the end with minimal supervision — at a reasonable price.

What you should know

Sonnet 5 also brings a few technical changes worth keeping in mind:

Pricing. The introductory price is $2 per million input and $10 per million output tokens, valid through August 31, 2026. After that it moves to $3 / $15. For comparison, Opus 4.8 costs $5 / $25. In Ayeto this is reflected in credits.
New tokenizer. The same text now produces roughly 30% more tokens than on Sonnet 4.6. The price per token doesn't change, but the real cost of a request may differ — factor that into your estimates.
1M-token context. Sonnet 5 ships with a one-million-token window by default, so it comfortably handles large documents and long conversations.
Safety safeguards. It's the first Sonnet with real-time protections for the cybersecurity domain. They're less strict than on Fable 5, but the model may decline risky requests.

How to switch it on in Ayeto

Ayeto's principle stays the same — no vendor lock-in. You assign Sonnet 5 like any other model:

Open the assistant you want to upgrade.
In the model settings, select Claude Sonnet 5.
Save and test it on a real task.

And if you're unsure whether Sonnet 5 is enough for a given task or whether it's worth reaching for Fable 5 or Opus, use Model Comparison and put them side by side — you'll see the difference in output, time, and credits.

Model information sources: Anthropic's official announcement and the Claude Platform documentation.

Try Sonnet 5 today — assign it to your most heavily used assistant and let us know how it did.