$2. That is the current price of a million input tokens on the Claude model Anthropic just made the default choice for every Free and Pro user. Claude Sonnet 5 launched on June 30, 2026, and it now sits at the center of the lineup: default for Free and Pro plans, available to Max, Team, and Enterprise users, live in Claude Code and on the Claude Platform, and callable through the API as claude-sonnet-5. It replaces Sonnet 4.6 as the default Sonnet-class model.
What changed with Sonnet 5
Anthropic built Sonnet 5 to be the most agentic Sonnet model yet. It plans multi-step work, uses tools such as browsers and terminals, and runs autonomously across longer stretches of a task without constant supervision. Anthropic states that its performance is close to that of Opus 4.8, the larger model in the same family, while running at a lower price. The company also reports a lower rate of undesirable behaviors than Sonnet 4.6 and describes Sonnet 5 as generally safer to use in agentic settings, so the extra autonomy comes with tighter guardrails than its predecessor rather than looser ones.
Where it sits in the family
Three models share the same lineage now, and the family keeps an unusually clean pricing pattern: every one of them charges exactly five times as much for output tokens as for input tokens.
- Claude Sonnet 5: $2 per million input tokens and $10 per million output tokens through August 31, 2026, then $3 and $15 afterward.
- Opus 4.8: $5 per million input tokens and $25 per million output tokens.
- Claude Fable 5: $10 per million input tokens and $50 per million output tokens. Anthropic took Fable 5 offline worldwide on June 13, 2026 under a US export control order, cleared it on June 30, and redeployed it worldwide on July 1 with a tightened safety classifier that blocks the original jailbreak technique in more than 99 percent of attempts, per [Anthropic's redeployment notice](https://www.anthropic.com/news/redeploying-fable-5). Pro, Max, Team, and select Enterprise plans include Fable 5 usage up to 50 percent of weekly limits at no extra cost through July 7, 2026; from July 8 it runs on paid usage credits at the rate above. The background is in ( today's Fable 5 update ).
The price gap, worked out
Run the ratios and Sonnet 5 comes out cheaper than Opus 4.8 on both meters, not on one at the expense of the other. At the introductory rate, Sonnet 5 costs 40 percent of the Opus 4.8 price on input and on output alike, a flat 60 percent discount either way. Once the regular rate takes over on September 1, Sonnet 5 settles at 60 percent of the Opus 4.8 price on both counts, a steady 40 percent discount rather than a cheaper input traded for a pricier output. Against Fable 5 the gap is wider again: Sonnet 5's regular price is 30 percent of Fable 5's rate on input and output alike, and the introductory price is 20 percent of it, a comparison that stays practical now that Fable 5 is back and billed at exactly that $10 and $50 rate once its included allowance runs out. For a model Anthropic itself places close to Opus 4.8 in capability, settling at three fifths of the cost on every token, in and out, is the number worth remembering once the introductory period ends.
The regular price is not a hike
The $3 and $15 rate that takes over on September 1 is not a new number introduced to soften a later increase. It is exactly what Sonnet 4.6 charged before it, so the introductory period works as a temporary discount layered onto an unchanged baseline rather than a low headline price that jumps once attention has moved on. Teams building against Sonnet 5 now can plan the September budget with confidence, because it is the same rate Sonnet has held for a full model generation.
Why the window before August 31 matters
For roughly two months, Sonnet 5 runs at 40 percent of the Opus 4.8 rate instead of the 60 percent it settles at afterward, on a model Anthropic positions as close to Opus in capability and more agentic than any Sonnet before it. Near-Opus output quality, full tool use, and a floor price that will not last make this window worth using for anything token-heavy: longer evaluation runs, bigger agentic pilots, or a plain cost comparison against whatever model is currently carrying production traffic. ( our $100 budget breakdown ) works through what a fixed budget actually buys at these rates. For the harder benchmark numbers behind the family's flagship model, ( our benchmark breakdown ) lines up the SWE-Bench Pro and analytical scores.
Bottom line
Sonnet 5 is not a side release. It is the model most Claude users are on by default as of June 30, 2026, priced below Opus 4.8 at every stage and a small fraction of the Fable 5 rate that now governs most Fable 5 usage once the free included allowance runs out. The introductory price closes on August 31; the regular rate waiting behind it is the same one Sonnet has charged for a generation. For ongoing coverage of the Claude model lineup, see ( the Fable 5 hub ).