Claude Token Counter: Opus 4.7 and Its Innovative Tokenizer
Le brief IA que les pros lisent chaque soir
Les 7 actus IA du jour, décryptées en 5 min. Gratuit.
Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.
Choisis ton rythme
Gratuit · Pas de spam · Désabonnement en 1 clic
An Enhanced Token Comparison Tool
The Claude Token Counter tool has been updated to allow for token count comparison between different Claude models. This new feature includes the Opus 4.7, Opus 4.6, Sonnet 4.6, and Haiku 4.5 models. While the Claude API accepts any model identifier, the main interest lies in comparing versions Opus 4.7 and 4.6, due to changes made to the tokenizer.
Opus 4.7: A Revamped Tokenizer
With the release of Opus 4.7, Anthropic has introduced an updated tokenizer. This new tokenizer improves text processing but results in an increase in the number of tokens for the same input. Depending on the type of content, this increase varies between 1.0 and 1.35 times. In a test using the system prompt of Opus 4.7, it was observed that the token count was 1.46 times higher than that of Opus 4.6.
Pricing Implications
Despite these changes, the pricing for Opus 4.7 remains the same as that of Opus 4.6, at $5 per million input tokens and $25 per million output tokens. However, the increase in the number of tokens could lead to a cost increase of about 40% for users.
Improvements for Image Processing
Opus 4.7 also offers better support for high-resolution images. It can now process images up to 2,576 pixels on the long side, which is approximately 3.75 megapixels, representing more than three times the capacity of previous models. In a test with a PNG image of 3456 × 2234 pixels weighing 3.7 MB, the number of tokens increased by 3.01 times compared to Opus 4.6.
Brief IA — L'actualité IA en français
L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.