Image: Anthropic / anthropic.com Anthropic Cancels the Sonnet 5 Price Hike — But Read the Tokenizer Note
The planned September 1 increase from $2/$10 to $3/$15 per million tokens is off. Sonnet 5 stays at introductory pricing permanently. The catch is a new tokenizer that produces about 30% more tokens for the same code.
The scheduled September 1 price increase for Claude Sonnet 5 is not happening.
Anthropic confirmed on its pricing documentation page that the $2/$10 per million input/output token rate — originally announced as introductory pricing through August 31 — is now the permanent price. The planned step-up to $3/$15 per million tokens is canceled.
For API users and teams building on Sonnet 5, that’s straightforwardly good news. The price that was supposed to expire tonight will hold indefinitely.
The tokenizer note that matters
Alongside the pricing freeze, the documentation carries a separate note worth reading: “Claude 4.7 and later models and Claude Mythos Preview use a newer tokenizer that contributes to their improved performance on a wide range of tasks. This tokenizer produces approximately 30% more tokens for the same text.”
Sonnet 5 falls under this. The per-token price is locked at $2 input and $10 output, but the same prompt that cost X tokens with Sonnet 4.6 will cost roughly 1.3X tokens with Sonnet 5 or later models.
For most conversational workloads the difference is modest. For code-heavy tasks — which tend to involve dense, structured text — the increase can push toward the higher end of that 30% range. Teams running large-scale code generation or code review pipelines should benchmark their actual token consumption before treating the price freeze as equivalent to a flat cost freeze.
How this compares
Sonnet 5 at $2/$10 with the new tokenizer still sits below Sonnet 4.6 at $3/$15 in per-token terms. The older model’s lower token count per prompt partially closes that gap, but Sonnet 5 will generally come out cheaper for most workloads — and meaningfully cheaper for anything requiring long context, since it includes the full 1M token window at standard pricing.
Competing models at similar price points include GPT-5.6 Luna and DeepSeek V4 Flash, both with lower list prices. Actual cost per task depends heavily on token efficiency per model, which varies by workload.
What prompted the freeze
Anthropic hasn’t detailed the business reasoning. The decision comes during a period when several major AI labs have revised pricing structures multiple times as compute costs shift and competition for developer adoption intensifies. Keeping Sonnet 5 at its launch price removes one friction point for teams evaluating whether to standardize on it.
The pricing page note is direct: “The previously scheduled increase to $3/$15 per million input/output tokens on September 1, 2026 will not occur.”
For teams that had built September budget projections around the higher rate, today’s confirmation means those plans can be revised down. The existing rate holds.