$2 / $10
per million tokens today: the introductory rate, in effect through August 31, 2026
Blog
Claude Sonnet 5 pricing: the $2/$10 rate ends August 31
Sonnet 5 is priced at $2 input / $10 output per million tokens, and that number has an expiry date most pricing pages skip: it is an introductory rate through August 31, 2026, after which it reverts to $3/$15. That is a scheduled 50% increase, and your forecast should use it.
Skip the read: get a measured recommendation in a few quick questions. Run the Stack Finder
$3 / $15
the rate from September 1: a 50% increase that is already on the schedule
$0.0410
our measured bill for a one-shot dashboard build: a fifth of Opus 5's price, and it worked first try
Claude Sonnet 5 costs $2 per million input tokens and $10 per million output through August 31, 2026. From September 1 the standard rate is $3/$15, a scheduled 50% increase. Cache reads bill at 10% of the input rate and the Batch API takes 50% off both sides.
How much does Claude Sonnet 5 cost?
Today: $2 per million input tokens, $10 per million output. From September 1, 2026: $3/$15 (pricing guide). Cache reads at 10% of input and the 50% Batch API discount apply at both rates. If you are modeling annual costs on the intro price, you are underestimating by a third for ten months of the year.
Want this priced against your own workload? Run the Stack Finder or start with $100 in credits.
What the deadline changes
At $2/$10, Sonnet 5 undercuts GPT-5.6 Terra ($2.50/$15) on both sides and is the obvious mid-tier. At $3/$15 it ties Terra exactly, and the decision moves to capability fit and ecosystem rather than price. Two practical moves before the deadline: benchmark your traffic on Sonnet 5 now while the testing is 33% cheaper, and if you are committing budgets, write September's numbers into the spreadsheet, not July's.
The same two tasks on every model we could meter
We sent an identical dashboard-generation prompt and an identical invoice-extraction prompt to every model we can call directly, on July 28, 2026, and recorded the actual bills. No estimates in the measured rows.
| Model | Rate $/M in / out | Dashboard task, measured | Extraction task, measured |
|---|---|---|---|
| Claude Fable 5 | $10 / $50 | $0.1462 | $0.0148* |
| Claude Opus 5 | $5 / $25 | $0.1158 | $0.0114 |
| Claude Opus 4.8 | $5 / $25 | $0.0724 | $0.0093 |
| Claude Sonnet 5 | $2 / $10 | $0.0410 | $0.0037 |
| Pareto | at cost | $0.0196 | $0.0018 |
| Claude Haiku 4.5 | $1 / $5 | $0.0162 | $0.0016 |
Measured 2026-07-28, one shot each, default settings, list prices. *Fable extraction is same-tokens list math from our earlier run; every other cell is a metered bill. Prompts published verbatim in the Fable 5 breakdown. All outputs completed the task; quality-per-dollar comparisons need the benchmark receipts, not this table alone.
Where Sonnet 5 actually sits
Our metered runs put its one-shot dashboard bill at $0.0410: 65% below Opus 5, 2.5x above Haiku, and the output worked first try. For extraction, summarization, classification, and most chat traffic, this is the tier where frontier headroom stops earning its premium. The catch is the same one every fixed-model choice carries: some fraction of your traffic will exceed it, and failed cheap runs bill you twice. The escalate-on-failure pattern, or a blended model that does it per request, is how the tier is used well.
Sonnet 5 pricing questions, answered
When does Sonnet 5's price go up?
September 1, 2026. The introductory $2/$10 runs through August 31, then the standard $3/$15 applies: a 50% increase on both sides that Anthropic scheduled at launch. Cache and batch discounts persist at the new rate.
Is Sonnet 5 cheaper than GPT-5.6 Terra?
Until August 31, clearly: $2/$10 versus $2.50/$15. From September 1 they tie at $3/$15 on... no: Terra stays $2.50/$15, so Terra becomes slightly cheaper on input while output ties. At that point pick on fit, not price.
Can Sonnet 5 replace Opus for coding?
For routine coding, often; our identical one-shot prompt produced a working build at a fifth of Opus 5's bill. For long agentic sessions and hard debugging, failure rates decide it: a cheap model that fails bills you for the retry on the expensive one anyway.
What happens to my costs if I lock in usage before September?
Nothing contractual: API pricing applies at request time, so August usage bills at $2/$10 and September usage at $3/$15 regardless of when you integrated. The intro window is for evaluation economics, not a lock-in.
Not sure which model fits?
The Stack Finder asks a few quick questions about your workload and gives you a straight recommendation. No account required.
Never think about tiers again. $100 to verify.
Buy Pareto API credits