Claude Sonnet 5 launched with introductory API pricing of $2 per million input tokens and $10 per million output tokens, scheduled to rise to $3/$15 — a 50% increase across the board — on 1 September 2026. On 10 August 2026 Anthropic cancelled the increase and made the introductory rate permanent.

What happened

Input (per 1M tokens)Output (per 1M tokens)
Introductory rate (from launch)$2.00$10.00
Planned standard rate, 1 Sep 2026$3.00$15.00
Actual rate from 10 Aug 2026$2.00$10.00

Anthropic’s own pricing documentation now records it directly: “The $2/$10 per million input/output token pricing for Claude Sonnet 5, announced at launch as introductory pricing through August 31, 2026, is now the standard price. The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur.”

A scheduled price rise being cancelled is unusual enough to be worth recording precisely, because plenty of secondary coverage described it as a “price freeze” or a “cut.” It was neither: the list price never changed. What changed is that a temporary rate became the standard rate, and a budgeted 50% increase disappeared. Batch API pricing follows at 50% of standard, so $1/$5 per million tokens.

Why it matters for builders

Delete the September increase from your 2026 forecast. If you built a cost model for Sonnet 5 workloads that steps up on 1 September — which was the correct thing to do at launch — that step is gone. For a workload running at, say, 500M input and 100M output tokens a month, the cancelled rise is roughly $1,000/month that will not now be spent. Check your FinOps assumptions and your reserved-budget alerts.

Inference pricing is now a competitive instrument, and it moves in both directions. Within the same month, Anthropic removed a scheduled increase and Google shipped Gemini 3.7 Flash at an introductory $0.75/$3.75 that expires on 31 December 2026 and doubles in January. The lesson is symmetrical: do not treat an introductory rate as permanent, and do not assume a scheduled increase will actually land. Both are marketing decisions taken against a competitor’s price sheet, not costs passed through.

Model your bill against list price, with a dated review. The defensible practice is to budget on the standard rate, record the date any introductory or promotional rate expires, and re-check before that date. Anything else leaves you exposed to a doubling you did not plan for — or, more happily, blind to a saving you already had.

Price stability is worth something on its own. Anthropic making a rate permanent, weeks before a rise it had already announced, is a signal aimed at enterprise buyers who need to commit to unit economics over a contract term. Read it alongside the competitive pressure on business users described in OpenAI’s enterprise crossover .

Sources

  1. Claude (Anthropic), announcement that Sonnet 5’s introductory pricing is permanent (10 August 2026): https://x.com/claudeai/status/2086891169217122586
  2. Anthropic, Claude platform pricing documentation (states the cancelled increase explicitly): https://platform.claude.com/docs/en/about-claude/pricing
  3. Anthropic, Claude pricing: https://claude.com/pricing
  4. Enterprise DNA, “Claude Sonnet 5 Price Freeze: What It Means for Business”: https://enterprisedna.co/resources/news/anthropic-claude-sonnet-5-pricing-permanent-reversal-august-2026/
  5. Big Hat Group, “Claude Weekly: Sonnet 5 Price Hike Canceled” (20 August 2026): https://www.bighatgroup.com/blog/claude-weekly-2026-08-20/

Further reading