The short version
- On June 30, 2026, Anthropic released Claude Sonnet 5, its most agentic Sonnet, and made it the default model on the Free and Pro plans.
- API price: $3 in and $15 out per million tokens, Anthropic's announced price after August 31, 2026. The launch price was $2 and $10.
- The same day, Fable 5 began returning to general availability after a US government directive suspended it on June 12.
- The advice: test Sonnet 5 on your highest-volume AI task, and keep a fallback for anything mission-critical.
What actually happened
On June 30, 2026, Anthropic released Claude Sonnet 5, its most agentic Sonnet so far. It can plan, use tools like browsers and terminals, and run tasks autonomously at a level that used to require larger, pricier models, and Anthropic puts its performance close to Opus 4.8. It is now the default model on the Free and Pro plans and is available to Max, Team, and Enterprise users.[1][2]
The same day, Claude Fable 5 began returning to general availability after a two-week ban. On June 12 the US government issued a directive to suspend access to Fable 5 and Mythos 5 by foreign nationals, after a method of jailbreaking the models surfaced. To comply, Anthropic switched both models off for all of its customers. With the security concern resolved, the directive was lifted. Mythos 5, the version without Fable's safeguards, is being redeployed first to a set of US organizations that defend critical infrastructure, while Fable 5 moves back toward general use.[4][5]
So in one day, the cheap end of the lineup got more capable, and the restricted top end came back from a government shutoff. Those two facts point in the same direction for anyone running real work on these models.
Sonnet 5 makes agents cheaper to run
The headline number is price. Sonnet 5 launched at an introductory $2 per million input tokens and $10 per million output through August 31, 2026. The announced price after that is $3 and $15.[3] The headline capability is agentic: this is a model built to run tools and multi-step tasks on its own. Near-Opus quality at 60 percent of Opus 4.8's price ($3 and $15 against $5 and $25 per million tokens) changes the math on what you can afford to automate.
Most of a business's AI work is routine, and routine work does not need the most expensive model. Every time a capable agentic model gets cheaper, more of that routine becomes worth automating, because the cost per task drops below the cost of a person doing it by hand. Sonnet 5 moves that line. The practical move is to take your highest-volume AI task and see how much of it Sonnet 5 handles well before you reach for a pricier model.
The Fable 5 ban is the part operators should not skip
It is easy to read the Fable 5 story as drama and move on. Do not. A frontier model your team might have been depending on was switched off for two weeks by a directive that had nothing to do with you, with no warning and no migration period. Access came back, but only because a government and a vendor worked it out.
While Fable 5 was unavailable, four open models responded before Anthropic could restore access.[6] A fallback has to be in place before the day you need it.
The lesson in one line
A cheaper, more capable model shipped on the same day a top model proved that access can be switched off overnight. Put the routine work on the model that does it well for less, and keep a second path for anything that must not stop.
Why the two on one day matter together
One end of the lineup got cheaper and more capable. The other end proved fragile. Together they make the same argument: do not architect anything you cannot afford to lose around one model staying available at one price.
Sonnet 5 looks like a good deal for high-volume work. Use it, and keep a fallback.
What to do this week
Three concrete moves:
- Test Sonnet 5 on your single highest-volume AI task and measure the cost difference against whatever you run now.
- Give anything mission-critical a second path: a cheaper or open model, or a second provider, that can take over if your primary is banned, throttled or repriced.
- Write down where your AI spend goes today, so you can plan for the next price change.
Common questions
How much does Claude Sonnet 5 cost?
Sonnet 5 launched at an introductory price of $2 per million input tokens and $10 per million output tokens through August 31, 2026. The announced price after that is $3 per million input and $15 per million output. It is the default model on Anthropic's Free and Pro plans and is available to Max, Team, and Enterprise users.
Is Claude Fable 5 available again?
Yes. After about two weeks the US government lifted the directive, and Fable 5 is returning to general availability. Mythos 5, the version without Fable's safeguards, is being redeployed first to a set of US organizations that operate and defend critical infrastructure, while Fable 5 moves back toward general use.
Should my business depend on a single AI model?
No. The Fable 5 ban showed that access to a frontier model can be switched off overnight by a policy change or a security incident. Keep a fallback, a cheaper or open model and ideally a second provider, so a single ban, outage, or price change cannot stop your operation.
If you want help
Tell us where your AI runs today. We will show you which work can move to a cheaper model, and where a single vendor is a single point of failure. Members of First Quarter get launches like this filtered to their own tools.
Sources
- Anthropic. "Introducing Claude Sonnet 5." anthropic.com, June 30, 2026.
- Wiggers, K. "Anthropic launches Claude Sonnet 5 as a cheaper way to run agents." TechCrunch, June 30, 2026.
- "Anthropic Launches Claude Sonnet 5 With Near-Opus Performance at a Lower Price." MacRumors, June 30, 2026.
- Anthropic. "Statement on the US government directive to suspend access to Fable 5 and Mythos 5." anthropic.com, June 2026.
- "US lifts ban on Anthropic's AI model Claude Fable 5 after security concerns." Crypto Briefing, June 2026.
- "Fable 5 ban: 4 open models responded before Anthropic could restore access." The New Stack, June 2026.
Update log
- Moved to the new Signals layout, trimmed our own commentary, noted that the launch price has ended, and stated the scope of the June 12 directive in one place.
- First posted.
Current as of June 30, 2026. Signals is general information, not legal, security or financial advice. How we source Signals.
