Claude Sonnet 5 launches with 1M token window, but tokenizer inflates costs

What happened

Anthropic released Claude Sonnet 5, a new AI model with performance close to its more expensive Opus 4.8 variant. The model has a 1 million token context window and 128,000 maximum output tokens. However, the new tokenizer produces approximately 30% more tokens than Claude Sonnet 4.6 for the same input, effectively raising the price by 30% despite list pricing remaining at $3/million input and $15/million output (with an introductory discount to $2/$10 through 31st August).

Why it matters

Sonnet 5 offers high performance at lower list prices, but the tokenizer change means real-world costs are higher than they appear. The model is less capable at cyber tasks than Anthropic's more powerful Mythos 5, which allowed the company to release it without US government restrictions. For developers and businesses using Claude, the effective price increase may offset the appeal of lower headline rates.

What to watch

The tokenizer efficiency varies by language—English text costs 1.42× more per token than Sonnet 4.6, Spanish 1.33×, Python code 1.27×, and Simplified Mandarin roughly the same. Adaptive thinking is enabled by default, and the model uses the same tools and platform features as Claude Sonnet 4.6.