Claude Sonnet 5, Explained Simply
What Claude Sonnet 5 is, what it does well, and what it costs โ released 2026-06-30 by Anthropic.
TLDR
- Claude Sonnet 5 is a newly released AI model from Anthropic.
- It can read huge amounts of text at once and "thinks" before answering by default.
- It has a cheaper introductory price, but uses a new way of counting words that might make it cost more in practice.
- It removes some hands-on technical controls that older models had.
Want a heads-up when Claude updates?
We email you the moment a new Claude version drops โ plus plain-English release notes like these. No spam.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Anthropic just released Claude Sonnet 5, a new AI language model. It is designed for developers and businesses who want to build apps that need to process massive amounts of text or write long, detailed responses.
What it does well
- Reads and writes a lot: It can handle up to 1 million "tokens" (pieces of words) at a time, which is like reading several thick books in one go. It can also generate up to 128,000 tokens in a single response.
- Thinks by default: It features "adaptive thinking" turned on automatically. This means the AI takes a moment to internally reason through a problem before giving you an answer, which usually leads to smarter results.
How it compares
- Compared to the previous model (Sonnet 4.6): It keeps the same general tools and features, but changes how it operates. The older model let users manually turn on "extended thinking" and adjust settings like "temperature" (which controls how random or creative the AI's answers are). Sonnet 5 removes those manual controls, choosing to handle the thinking process automatically instead.
What it costs and what it can handle
- Price: It has an introductory price of $2 per million tokens for reading (input) and $10 per million tokens for writing (output). This deal lasts until August 31, 2026. After that, the standard price goes up to $3 and $15.
- Limits: It can read up to 1 million tokens at once and write up to 128,000 tokens in one reply.
Worth knowing
- Word counting change: It uses a new "tokenizer" (the tool that breaks text into countable pieces). This new tool creates about 30% more tokens for the exact same text. Since you pay by the token, your bills might be a bit higher than expected.
- Missing features: It does not offer a "Priority Tier" (a paid upgrade for faster service speeds).
- Strict rules for developers: If a programmer tries to use the old manual thinking settings or changes the "temperature" settings to make the AI more creative, the system will completely reject the request with a "400 error."
This is our plain-language summary. Read the complete, official notes on the Claude official changelog โ.
