Short answer: you have until August 31, 2026. Claude Sonnet 5 launched June 30 as the default model on every Claude plan at an intro rate of $2/$10 per million tokens — on September 1 it rises to $3/$15, a 50% increase. The detail most coverage undersells: its new tokenizer produces about 30% more tokens for the same input text (Anthropic's own published figure). Six weeks of cheap headroom to benchmark and migrate. Checklist below.
What Sonnet 5 actually is (60 seconds)
Sonnet 5 is Anthropic's everyday-tier model and the new default everywhere — Free plan, Pro, claude.ai, Claude Code, and Cowork all switched on launch day, June 30. It sits in a four-rung ladder: Haiku 4.5 ($1/$5) → Sonnet 5 ($2/$10 intro → $3/$15 standard) → Opus 4.8 ($5/$25) → Fable 5 ($10/$50). Anthropic's pitch: near-Opus coding and agentic quality at Sonnet cost, with the largest capability gains over Sonnet 4.6 concentrated in coding and agentic tasks (Anthropic's Transparency Hub has the full benchmark breakdown).
The two numbers that decide your bill
- Intro pricing expires. $2/$10 becomes $3/$15 on September 1 — the standard rate identical to Sonnet 4.6's per-token price. Anthropic frames the intro window as making migration "roughly cost-neutral" against that baseline.
- The tokenizer tax. Sonnet 5 ships a new tokenizer that Anthropic's own documentation says produces approximately 30% more tokens for the same input text than Sonnet 4.6 — the exact increase depends on the content. Per-token pricing is unchanged, but because the same text now costs more tokens, the real cost of an equivalent request can differ from Sonnet 4.6 even before the September price rise. Recount your prompts against Sonnet 5 specifically; don't reuse old token counts.
The levers that survive the price change: cached input at roughly a tenth of list and Batch API at 50% off. In multi-turn agent loops, where full history is resent every turn, caching isn't an optimization — it's the difference between viable and not.
The solo-operator migration checklist (use the cheap weeks)
- Benchmark your tokenizer impact now. Run your 10 most common prompts through Sonnet 5's token counter and compare against your current model. Text-heavy workloads (blog drafts, briefings) carry more risk than code.
- Set up prompt caching before September. Same system prompt resent daily? That's cached-input pricing instead of full-rate input.
- Route volume to Haiku 4.5 ($1/$5). Classification, extraction, repurposing — Haiku handles the grind at a fraction of Sonnet's standard rate.
- Keep Sonnet 5 for production work. Customer-facing copy, briefings, agentic tasks — where the quality jump justifies the price.
- If you're on Cowork already, you're on Sonnet 5. It became the Cowork default at launch — check your usage this week, not in September.
How it stacks against the July alternatives
| Model | Input / Output per 1M | Best for |
|---|---|---|
| Sonnet 5 (intro) | $2 / $10 — until Aug 31 | Agentic work, production copy |
| Sonnet 5 (standard) | $3 / $15 — from Sep 1 | same, pricier |
| GPT-5.6 Terra | $2.50 / $15 | OpenAI-ecosystem equivalent |
| Grok 4.5 | $2 / $6 | Agentic coding (not yet in EU) |
| Haiku 4.5 | $1 / $5 | Volume/grind tasks |
At intro rates Sonnet 5 is strong value in its class. At standard rates the picture tightens — Grok 4.5 undercuts it on list output price (if you're outside the EU and its vendor-reported efficiency claims hold on your workload — list price alone doesn't decide the winner; measure per task, not per token).
Who doesn't need to act
- Free/Pro consumer users: Sonnet 5 is already your default at no extra cost — the price change is API-only.
- Light users under roughly 1M tokens/month: the difference is small money. Focus elsewhere.
- Anyone already caching + batching: your effective rate barely moves.
FAQ
- When does Claude Sonnet 5's introductory pricing end?
August 31, 2026. From September 1, API pricing rises from $2/$10 to $3/$15 per million tokens. Consumer plan pricing (Free/Pro/Max) is unaffected.
- Is Sonnet 5 the default Claude model now?
Yes — since June 30, 2026 it is the default on the Free plan, Pro, claude.ai, Claude Code, and Cowork. If you used Claude at all this month, you likely already used Sonnet 5.
- Does Sonnet 5 really use more tokens for the same text?
Yes. Anthropic's own documentation states the new tokenizer produces approximately 30% more tokens for the same input text than Sonnet 4.6; the exact increase depends on the content. Text-heavy workloads should recount tokens before the September price rise.
Pricing and tokenizer figures verified July 22, 2026 against Anthropic's official "What's new in Claude Sonnet 5" documentation. Daily operator signals: BEARTH SIGNAL · Every workflow as a deployable template: BEARTH VAULT