Claude Sonnet 5 Pricing: What the Cheaper Agent Model Really Costs

Key takeaways

• Sonnet 5 introductory pricing is $2/M input and $10/M output, rising to $3/M and $15/M after August 31, 2026.
• Opus 4.8 costs $5/M input and $25/M output, so Sonnet 5 lands at roughly 40% of the flagship's price.
• A new tokenizer can map the same text to up to 1.35x more tokens, which offsets part of the sticker discount.
• Anthropic set the intro price so switching stays roughly cost-neutral versus the prior Sonnet on typical text.
• For agents that run all day, output-token volume, not the headline rate, drives the bill.

What is Claude Sonnet 5 and why does pricing matter?

Anthropic released Claude Sonnet 5 on June 30, 2026, positioning it as a mid-tier model that runs close to the flagship Opus 4.8 on reasoning, coding, and tool use while costing less than half as much. The pitch is simple: make running agents cheap enough to keep them working all day. For anyone budgeting agentic workloads, the launch resets the cost math.

How much does Claude Sonnet 5 cost?

Introductory pricing is $2 per million input tokens and $10 per million output tokens, in effect through August 31, 2026. After that, the rate moves to $3 per million input and $15 per million output. By comparison, Opus 4.8 sits at $5 input and $25 output. On the sticker, Sonnet 5 is about 60% cheaper than Opus during the intro window and roughly 40% cheaper once standard pricing applies.

Why is the new tokenizer a catch?

Sonnet 5 uses a new tokenizer, so the same text can map to up to 1.35x more tokens than before. That matters because you pay per token, not per word. A 35% increase in token count quietly eats into the headline discount. Anthropic says it set the introductory price so the switch stays roughly cost-neutral against the previous Sonnet on typical text, which is a tell: the sticker price and the effective price are not the same number. Model your own prompts to see where you actually land.

How do you model the real cost of an agent?

Agents are output-heavy. They plan, call tools, read results, and write again, often looping many times per task. Output tokens cost five times more than input on Sonnet 5 ($10 vs $2 at intro pricing), so a chatty agent that emits long reasoning traces skews the bill toward the expensive side of the ledger. For example, say a single run consumes 50K input and 20K output tokens. At intro pricing that is (0.05 x $2) + (0.02 x $10) = $0.30 per run. Multiply by daily run volume, add the tokenizer inflation, then check what happens on September 1 when the rate steps up. The headline "cheaper" claim is true, but the size of the saving is yours to calculate, not Anthropic's to promise. You can model Sonnet 5 against Opus 4.8 side by side in Calcaas before you commit a workload.

Where can you use Sonnet 5?

Sonnet 5 is the default model for Claude Free and Pro users immediately, and is available to Max, Team, and Enterprise customers. It also ships in Claude Code and on the Claude Platform, so the same pricing applies whether you are prototyping in an IDE or running production agents through the API.

Takeaway: Sonnet 5 is genuinely cheaper than Opus, but the tokenizer change and the August 31 price step mean you should model your actual workload before assuming the savings.