Core Insights - Claude Opus 4.6 has launched a Fast Mode that increases speed to 2.5 times that of the normal mode while maintaining performance consistency [1] - The pricing for Fast Mode is significantly higher, costing $150 per million tokens for output compared to $25 for the standard mode, marking a sixfold increase [4][9] - Fast Mode operates independently of subscription limits, meaning users will incur additional charges from the first token used in this mode [5] Pricing and Cost Structure - Standard API pricing for Claude Opus 4.6 is $5 per million tokens for input and $25 for output, making it one of the most expensive models in the market [6] - Fast Mode pricing is set at $30 per million tokens for input and $150 for output, with costs nearly doubling for long context scenarios [9] - Users have expressed concerns that such high pricing may lead to a loss of customers for Anthropic [11][12] Performance and Capabilities - Despite the higher costs, the model quality remains the same in Fast Mode, with no enhancements in intelligence or response quality [5] - Opus 4.6 has been recognized as the top-performing model in various assessments, including the Intelligence Index v4.0, where it scored 53 points, surpassing OpenAI's GPT-5.2 by two points [15] - The model has shown significant improvements in coding and reasoning tasks, with a notable increase in scores across various benchmarks [24][25] Context and Self-Correction Features - Opus 4.6 supports a context window of up to 1 million tokens, a substantial increase from the previous limit of 200,000 tokens, which helps mitigate context decay issues [26] - The model exhibits advanced self-correction capabilities, allowing it to autonomously assess task difficulty and adjust its approach accordingly [29] Market Implications - The introduction of Fast Mode signals a shift in the AI industry, emphasizing the importance of speed alongside intelligence [34][36] - The pricing strategy reflects a market test by Anthropic to gauge how much users are willing to pay for increased speed in AI operations [38]