On October 7, 2026, Anthropic released Claude Haiku 5.5, the latest version of its small-scale AI model, now available on the Claude platform as well as cloud services like Amazon Web Services, Google Cloud, and Microsoft Azure. The company claims that the operating cost of this model is about 75 percent lower than that of the previous version, Haiku 4.5. Anthropic describes Haiku 5.5 as the most affordable, fastest, and most capable small model it has developed, optimized for high-volume, cost-sensitive tasks such as summarizing text, compressing data, querying databases, and classifying information. The model can also function as a sub-agent alongside Claude Opus 5.5 and Claude Sonnet 5.5 to assist with coding tasks and is suitable for applications where speed is critical, such as live customer support and automated browser interactions. Anthropic notes that Haiku 5.5 is the first of its kind to include an adjustable effort parameter, allowing users to prioritize either cost efficiency or the model's intelligence. Developers can access the model using the identifier "claude-haiku-5-5," and the company has provided a migration guide for the update. According to technical specifications, Haiku 5.5 was trained using a mix of publicly available internet data, public and private datasets, user data authorized for training, and synthetic data. Its knowledge cutoff date is set to June 2026, and it is designed to generate only text-based responses. Pricing for Haiku 5.5 is significantly lower than for previous models. For requests under 100,000 tokens, the cost is $0.10 per million input tokens and $0.50 per million output tokens, with cached reads and writes costing $0.01 and $0.125, respectively. For larger requests, the rates increase to $0.50 for input, $2.50 for output, $0.05 for cached reads, and $0.625 for cached writes per million tokens. In comparison, Haiku 4.5 had input costs of $1.00 and output costs of $5.00 per million tokens, with higher rates for cached operations. A footnote in the announcement explains that the 75 percent cost reduction is based on a 90 percent price drop for smaller requests and a 50 percent drop for larger ones, with most Haiku 4.5 requests falling into the lower pricing tier. This change also accounts for an updated tokenizer, similar to those in Sonnet 5.5 and Opus 5.5, which slightly increases token usage. This announcement follows a report indicating that Anthropic’s Claude subscriptions offer significantly more value than OpenAI's ChatGPT, with a study by SemiAnalysis suggesting that Claude provides about five times the API value of ChatGPT. However, despite subsidies from AI companies, the high cost of computing resources continues to challenge profitability, making budget control difficult for large organizations. Some companies are turning to hybrid strategies to manage expenses. Meanwhile, Anthropic plans to caution potential investors during its initial public offering (IPO) about the potential "catastrophic or existential risks" posed by advanced AI, a rare warning from a company seeking to profit from the same technology. The company, which positions itself as a private AI laboratory focused on safety, has dedicated nearly 80 pages of its 261-page prospectus to risk factors, nearly double the pages allocated to its business description. Anthropic aims for a valuation of $200 billion, despite reporting rising losses, high investment costs, and ongoing security challenges.