Claude Sonnet 5 Pricing:
$2 / $10 is now the standard rate
The complete cost-per-token breakdown for Claude Sonnet 5 (released June 30, 2026): the standard $2/$10 rate, prompt-caching rates, and the new-tokenizer caveat that changes the real math. The increase once scheduled for September 1, 2026 has been cancelled.
Update: Claude Sonnet 5's $2/$10 was confirmed as the standard rate on 2026-08-10. The increase to $3/$15 once planned for 2026-09-01 will not happen.
Introductory vs standard rates
Claude Sonnet 5 launched with promotional pricing that was scheduled to end on August 31, 2026. On August 10, 2026 Anthropic confirmed that this rate became the standard price and that the September 1 increase to $3/$15 will not occur. The figures below are therefore the long-term prices.
Introductory pricing
Standard rate (current)Promotional launch pricing for all Claude Sonnet 5 usage until the intro window closes. Best treated as roughly cost-neutral versus Sonnet 4.6, not a permanent discount (see the effective-cost note below).
Standard pricing
Planned increase · cancelledThe ongoing rate once the introductory period ends. Plan any longer-running or recurring workload against these numbers so budgets stay accurate past August.
Prices are Anthropic's published list rates for claude-sonnet-5, in USD per million tokens (MTok). Source: anthropic.com/news/claude-sonnet-5 and platform.claude.com docs.
Prompt-caching rates
Prompt caching cuts the cost of repeated context. Cache reads scale with the current input tier (intro or standard), while cache writes are priced as a multiple of the base input rate.
| Cache operation | Introductory | Standard | Notes |
|---|---|---|---|
| Cache read | $0.20 / MTok | $0.30 / MTok | Reusing cached tokens |
| 5-minute cache write | 1.25x | 1.25x | 1.25x base input rate |
| 1-hour cache write | 2x | 2x | 2x base input rate |
Cache-write multipliers apply to whichever base input rate is active ($2/MTok intro or $3/MTok standard). A 5-minute write is 1.25x that base; a 1-hour write is 2x. Cache reads fall to $0.20/MTok during the intro period and $0.30/MTok at standard pricing.
The real effective cost
The sticker price only tells half the story. Claude Sonnet 5 ships a new tokenizer, and that changes how many tokens the same text consumes.
The new tokenizer uses ~30% more tokens
The same text is encoded into roughly 30% more tokens under Claude Sonnet 5's tokenizer than under Sonnet 4.6. Because billing is per token, that offsets most of the lower sticker price. The $2/$10 rate is best described as roughly cost-neutral versus Sonnet 4.6's $3/$15 on identical text, not a flat 33% discount. Always estimate against your own real token counts.
What looks cheaper
Per token, Sonnet 5's $2/$10 sits about a third below Sonnet 4.6's $3/$15. That rate is now confirmed as the standard price rather than a time-limited promotion.
What actually happens
Because each request carries ~30% more tokens, the effective per-request cost is higher than the sticker delta implies. Net effect: intro pricing lands close to cost-neutral versus Sonnet 4.6 for the same content.
Rate-card comparison
How Claude Sonnet 5 sits against the models it slots between and above in Anthropic's line-up. Sonnet 5 is positioned mid-tier, between Claude Haiku 4.5 and Claude Opus 4.8.
| Model | Input | Output | Context | Positioning |
|---|---|---|---|---|
| Claude Sonnet 5 | $3 / MTok | $15 / MTok | 1M | Mid-tier, standard rate |
| Claude Opus 4.8 | $5 / MTok | $25 / MTok | 1M | Frontier / hardest tasks |
| Fable 5 | $10 / MTok | $50 / MTok | 1M | Premium tier |
Sonnet 5 figures shown are the standard $2/$10 rate, confirmed as the long-term price on 2026-08-10. Anthropic describes Sonnet 5 as approaching Opus 4.8 quality at roughly 40% lower price, but Opus 4.8 still leads the hardest coding, judgment, and cyber tasks.
How QCode billing works for Sonnet 5
QCode does not invent its own Sonnet 5 rate. Usage is charged proportionally to Anthropic's official list price and drawn from a single balance that spans every model.
Proportional to official price
Your Sonnet 5 usage tracks Anthropic's published rate of $2/$10. Because the September 1 increase was cancelled, that rate does not change on that date.
One balance across all models
Sonnet 5, Opus 4.8, Fable 5 and every other model share the same QCode balance. Mix models freely in one project with no separate top-up or per-model quota to manage.
Cost model
Because Sonnet 5's tokenizer produces more tokens per request, watch your real token counts rather than word or character estimates when forecasting spend.
Update: the introductory rate became standard and no longer expires
Introductory pricing is a launch promotion with a hard end date. There is no auto-extension. If you are budgeting a workload that runs into the autumn, model it on the standard rate so there are no surprises when the switch happens.
Frequently asked questions
How much does Claude Sonnet 5 cost per token?
Claude Sonnet 5 is $2 per million input tokens and $10 per million output tokens, and that is the standard rate. Cache reads are $0.20/MTok, a 5-minute cache write costs 1.25x base input, and a 1-hour cache write costs 2x base input. The increase to $3/$15 once scheduled for September 1, 2026 was cancelled on August 10, 2026.
When does Claude Sonnet 5 introductory pricing end?
No longer. The $2/$10 per-MTok rate began as a launch promotion scheduled to end on August 31, 2026. On August 10, 2026 Anthropic confirmed it had become the standard price and that the September 1 increase would not occur, so you can plan long-running workloads against $2/$10 directly.
Is Claude Sonnet 5 cheaper than Claude Opus 4.8?
Yes. Claude Opus 4.8 is $5 per million input tokens and $25 per million output tokens. At Sonnet 5's $2/$10 that is roughly 60% cheaper per token. Anthropic positions Sonnet 5 as approaching Opus 4.8 quality at a lower price, though Opus 4.8 still leads the hardest coding and judgment tasks.
Why does Claude Sonnet 5 use more tokens for the same text?
Claude Sonnet 5 ships a new tokenizer that encodes the same text into roughly 30% more tokens than Claude Sonnet 4.6. Because you are billed per token, that offsets much of the lower sticker price: the $2/$10 rate is best described as roughly cost-neutral versus Sonnet 4.6's $3/$15 on identical text, not a flat 33% discount. Always estimate against your own real token counts.
How does QCode bill for Claude Sonnet 5?
QCode charges proportionally to Anthropic's official Sonnet 5 rate and draws from one shared balance that works across every model. The current rate is $2/$10 and it does not change on September 1. You can mix Sonnet 5, Opus 4.8, Fable 5 and other models on the same balance with no separate top-up.
Is Claude Sonnet 4.6 being retired now that Sonnet 5 is out?
No. Claude Sonnet 4.6 (claude-sonnet-4-6) remains Active with a tentative retirement no sooner than February 17, 2027. Sonnet 5 is the recommended successor and the new default, but there is no forced-migration deadline, so you can keep running Sonnet 4.6 while you evaluate the switch.
Run Claude Sonnet 5 on QCode
One balance, every model, pricing that tracks Anthropic's official rates. Start with Sonnet 5 today.