A website has emerged that visualizes how much token consumption has increased with Claude Opus 4.7, showing examples where the same input consumes twice as much as in version 4.6.
The '
Tokenomics - Anthropic Token Cost Calculator
https://tokens.billchambers.me/
Claude Opus 4.7 is a model with improved coding and visual understanding capabilities compared to Claude Opus 4.6. While the API usage fees remain the same as Claude Opus 4.6, Anthropic has stated that token consumption may increase for the same prompt due to changes in the tokenizer. AI models like Claude tokenize user-inputted prompts using a tokenizer before processing them, and the API usage fees for AI services are determined by the number of tokens input and output. In other words, while the cost per token for Claude Opus 4.7 is the same as Claude Opus 4.6, the usage fee may increase for the same prompt because it requires more tokens.
Claude Opus 4.7 Released: Improved Instruction Tracking and Image Recognition Capabilities, while Maintaining the Same Price as Opus 4.6 - GIGAZINE
Tokenomics is a web application released by AI expert Bill Chambers following changes to the tokenizer in Claude Opus 4.7, which allows you to see the difference in token consumption when entering the same prompt in Claude Opus 4.7 and Claude Opus 4.6.
Go to Tokenomics, enter the prompt, and then click 'Calculate Request Difference'. This time, following the example posted by Chambers, I entered 'WHO ARE YOU'.
The results of calculating the number of tokens consumed and the price using Anthropic's
When typing 'who are you' in lowercase, the number of tokens consumed remains 10 in Claude Opus 4.6, but it has decreased to 14 in Claude Opus 4.7. Even so, Claude Opus 4.7 still consumes 1.4 times more tokens.
Tokenomics also provides a results summary page for all users . According to the summary page, Claude Opus 4.7 increased the number of tokens by an average of 38.6%. At the bottom of the summary page, recent input results are displayed, clearly showing that Claude Opus 4.7 consumes more tokens regardless of whether the text is short or long.
There is other information indicating an increase in token consumption in Claude Opus 4.7. The following link shows verification results stating that 'token consumption in Claude Opus 4.7 has increased by 1.005 to 1.07 times for Japanese, Chinese, Korean, and emojis, and by 1.20 to 1.47 times for English and code compared to Claude Opus 4.6.'
I Measured Claude 4.7's New Tokenizer. Here's What It Costs You.
Additionally, comments such as 'I'm subscribed to the Pro plan, but I reached my usage limit after only four attempts to create a webpage' and 'I'm subscribed to the Max 5x plan, but I reached my 5-hour usage limit in just 2 hours' have been posted on the news sharing site Hacker News .
Furthermore, an analysis by the third-party organization Artificial Analysis found that 'while input costs for Claude Opus 4.7 have increased compared to Claude Opus 4.6, inference and output costs have decreased, resulting in an overall cost that is slightly lower than Claude Opus 4.6.' Therefore, for workloads that heavily utilize inference, Claude Opus 4.7 may be cheaper.
Related Posts:
in AI, Posted by log1o_hf