Claude Sonnet 5: What's New and How to Use It

Claude Sonnet 5 is Anthropic's most agentic Sonnet model, with performance close to Opus 4.8 at a lower price, a 1M-token context window, and free access on Claude's consumer plan. Anthropic launched it on June 30, 2026 and made it the default model for the Free and Pro plans, so it is the model most people now mean when they say "Claude."

If you are trying to figure out what Claude Sonnet 5 actually changes, whether you can use it for free, and how to put it to work, this is the short version and the practical details.

Key Takeaways:

Claude Sonnet 5 is a big step up over Sonnet 4.6 in reasoning, tool use, and coding, with agentic performance close to Opus 4.8
It has a 1M-token context window, is the default model on Claude's Free and Pro plans, and adds real-time cybersecurity safeguards
Yes, it is free at the consumer tier: Sonnet 5 is the default model on Claude.ai's Free plan
A new tokenizer produces about 30% more tokens for the same text, so budget by effective cost, not the rate card alone

What Claude Sonnet 5 Is

Claude Sonnet 5 is the next generation of Anthropic's mid-tier Sonnet family, positioned as a cheaper way to run agents than the flagship Opus line. The pitch is most of Opus 4.8's capability at a fraction of the cost, which is why it became the default model across Claude's plans at launch.

The word that matters is agentic. Sonnet 5 is built to make plans, use tools like browsers and terminals, and run multi-step work autonomously, not just answer a question. That is the difference between a chat model and a model you can hand a task to.

What this resolved: Sonnet 5 is the new default Claude, tuned for agent work. What it left unsolved: "agentic" only pays off if the workflow around the model is built for it.

What's New in Claude Sonnet 5

Four changes matter most, per Anthropic's launch notes:

More agentic. Clear gains over Sonnet 4.6 in reasoning, tool use, and coding, landing close to Opus 4.8 on agentic tasks.
1M-token context window. One million tokens by default, enough to hold a large codebase, a long document set, or a full research corpus in a single pass.
Real-time cybersecurity safeguards. The first Sonnet-tier model with them; flagged requests return a refusal rather than an unsafe answer.
A new tokenizer. The same text now produces roughly 30% more tokens than on Sonnet 4.6, with per-token pricing unchanged. That raises the effective cost of an equivalent request even though the rate card did not change. We break down the math in our guide to the Sonnet 5 tokenizer cost change.

There are also three API behavior changes worth knowing if you build on it: adaptive thinking is on by default, manual extended thinking now returns a 400 error, and non-default sampling parameters return a 400. Those trip up migrations from Sonnet 4.6.

What this resolved: the real upgrades are agentic ability, context size, safety, and a denser tokenizer. What it left unsolved: the tokenizer change quietly shifts your cost, covered next.

Is Claude Sonnet 5 Free

Yes, at the consumer tier. Claude Sonnet 5 is the default model on Claude.ai's Free plan, so you can use it at no cost for chat-style work: questions, drafting, analysis, and light coding help. Anthropic made it the free-plan default at launch, which is unusual for a model this capable.

The honest caveats: the Free plan has usage limits and gives you the chat product, not automated agent workflows or guaranteed throughput. For heavier or automated use, you move to Pro, the API (pay-as-you-go), or an agent workspace. But if your question is simply "can I try Claude Sonnet 5 without paying," the answer is yes.

What this resolved: free access to the model genuinely exists on Claude.ai. What it left unsolved: free means the chat model with limits, not automated or high-volume use.

How to Use Claude Sonnet 5

There are three practical paths, by how much you need.

Claude.ai Free or Pro. Sign in and Sonnet 5 is the default. Best for chat, drafting, and one-off tasks. Free has limits; Pro raises them.
The Claude API. Pay-as-you-go with full control over prompts, tools, and integration. It runs $2/$10 per million input/output tokens under introductory pricing through August 31, 2026, then $3/$15. This is the developer path.
Inside an agent workspace. If you want Sonnet 5 wrapped in research-to-output workflows without touching an API key, MoClaw runs Claude Sonnet 5 inside a managed cloud workspace, available now. You describe the outcome and the agent researches, structures, and produces the file.

Ravi, a consultant who is not a developer, hit the Free plan's ceiling mid-project and did not want to manage API billing. Running Sonnet 5 through a MoClaw agent let him hand off research-and-draft tasks and get finished documents back, which the chat-only free tier could not do.

What this resolved: a free-to-API-to-agent ladder you can pick from. What it left unsolved: the right rung depends on whether you need chat, code control, or hands-off output.

Claude's Free Plan vs Sonnet 5 in an Agent Workspace

A common point of confusion: is running Sonnet 5 in MoClaw the same as using Claude.ai? No. It is the same model, wrapped in a different job. MoClaw runs Claude Sonnet 5 through Anthropic's API as an integration; it does not claim to be Claude.

	Claude.ai Free plan	Sonnet 5 in MoClaw
The model	Claude Sonnet 5	Claude Sonnet 5 (via integration)
What you get	Chat responses	Agent workflows: research to output
Produces files	Copy from chat	Finished docs, decks, data feeds
Runs recurring jobs	No	Yes

The difference is not the model. It is the workflow around it: research to a document or deck, memory across runs, and delivery where you work.

What this resolved: same model, different unit of work (chat vs finished output). What it left unsolved: which you need depends on whether you want answers or artifacts.

FAQ

Is Claude Sonnet 5 free?

Yes, at the consumer tier. It is the default model on Claude.ai's Free plan, which has usage limits. Heavier or automated use moves to Pro, the API, or an agent workspace.

What is new in Claude Sonnet 5 versus Sonnet 4.6?

Stronger agentic performance (reasoning, tool use, coding), a 1M-token context window, real-time cybersecurity safeguards, and a new tokenizer that produces about 30% more tokens for the same text.

Do I need an API key to use Claude Sonnet 5?

Not for Claude.ai or for running it inside MoClaw. You only need an API key if you build directly on Anthropic's API.

How much does Claude Sonnet 5 cost?

Introductory API pricing is $2/$10 per million input/output tokens through August 31, 2026, then $3/$15. The consumer chat tier is free with limits. See our Claude Sonnet 5 pricing breakdown.

Where can I use Claude Sonnet 5?

On Claude.ai (free and paid), via the Claude API, through AWS, Google Cloud, and Microsoft Foundry, and inside agent workspaces like MoClaw.

Getting the Most Out of Claude Sonnet 5

Claude Sonnet 5 is the right default for most work in 2026: close to Opus on capability, free to try, and built for multi-step tasks. Start on the free tier to see how it handles your work, and move to the API or an agent workspace when you need automation, files, or higher throughput.

If you want Sonnet 5's capability without managing keys, budgets, or migration traps, MoClaw runs it behind a plain-English agent, available now. Describe what you need and let the agent do the multi-step work, then read our pricing breakdown before you scale so the tokenizer change does not surprise your bill.

Editor's note: Claude Sonnet 5 details were verified against Anthropic's official documentation on July 10, 2026. Pricing, availability, and plan defaults can change; check the linked primary sources before making production decisions. MoClaw runs Claude Sonnet 5 via Anthropic's API and is not affiliated with or endorsed by Anthropic.