Future TechnologyClaude Sonnet 5 Is Now the Default Model for Every Free and Pro User
Key takeaways
- Claude Sonnet 5 launched June 30 and became the default model for every Free and Pro user on July 1
- Anthropic says it performs close to flagship Opus 4.8 on many agentic tasks, including planning, browser use, and terminal work
- Pricing is discounted through August 31, undercutting Sonnet 4.6's launch price
- California's state government separately rolled out the largest US state AI deployment yet, giving agencies and opted-in cities and counties Claude access at 50% off
Open a free Claude account today and you're talking to a model that, a year ago, would have needed a paid tier and a much bigger parameter count to keep up. Anthropic launched Claude Sonnet 5 on June 30 and flipped it to the default model for every Free and Pro user the very next day, no upgrade required.
What actually changed
Anthropic is calling Sonnet 5 the most agentic Sonnet it has built. That's not marketing filler if you look at what it does differently: it plans multi-step work on its own, calls a browser and a terminal without being walked through every click, and checks its own output before handing back a result. The company says performance on a lot of these tasks now sits close to Opus 4.8, the flagship model that used to be the only option for anything demanding.
The pricing move matters as much as the capability jump. Sonnet 5 is discounted through August 31, undercutting where Sonnet 4.6 launched. For a Free or Pro user, that's a better model at a lower cost, landing in your account with no action needed.
The bigger rollout nobody expected
The same week, California's state government rolled out what's being described as the largest AI deployment by a US state government to date. Every state agency gets access, and any city or county that opts in can use Claude through a new shared-services portal at a 50% discount. That's not a pilot program, it's procurement at a scale that other states will be watching closely before deciding whether to follow.
Put the two stories together and a pattern shows up. This follows Nvidia's Rubin platform push to cut the cost per AI token by up to 10x, and it lands in the same stretch of talent moves we covered when Noam Shazeer left Google for OpenAI. Cheaper inference, bigger model capability, and government-scale adoption are arriving in the same few weeks rather than one at a time.
Why it matters
The AI assistant race has been framed for two years as "who has the smartest model." That framing is getting harder to defend. The more useful question now is whose model can actually finish a task unsupervised, and whose employer or government is willing to bet real budget on that happening reliably. Free users getting agentic capability by default, with no opt-in, resets what people expect from a chatbot the next time they open one.
It also raises the stakes on trust. An assistant that books things, edits files, and runs commands on your behalf needs to be right more often than one that just answers questions. Anthropic's bet is that Sonnet 5 clears that bar widely enough, at a low enough price, that the trade-off becomes worth it for millions of people who never asked for an agent in the first place.
What's driving the timing
If you're curious why this is arriving now rather than three years ago, Samsung's $648 billion bet on the chip manufacturing behind this wave of AI is the infrastructure story underneath it. Cheaper, faster chips are what make agentic models affordable enough to hand out for free.