Anthropic, a company specializing in artificial intelligence, has released a new version of its Claude Haiku model, called Claude Haiku 5.5. This model is the smallest in the Claude family and is designed for tasks that are simple and repetitive, such as summarizing documents or sorting requests. On Wednesday, October 7, Anthropic announced the release of Claude Haiku 5.5, which comes with a new pricing structure for developers who pay for AI usage. The cost for using Claude Haiku 5.5 is $0.10 (about €0.09 HT) per million input tokens and $0.50 (about €0.45 HT) per million output tokens. This represents a significant decrease compared to Claude Haiku 4.5, which was released in October 2025 and cost $1 and $5 per million tokens for input and output, respectively. A token is a unit of text, typically a fragment of a word or a punctuation mark, used by AI providers to measure and bill for usage. For requests under 100,000 tokens, Claude Haiku 5.5 is ten times cheaper than Claude Haiku 4.5. Anthropic describes Claude Haiku 5.5 as the fastest model in the Haiku series and the first to offer an effort adjustment, allowing developers to choose between a more thoughtful response and a cheaper one. The model is intended for tasks that are short and numerous, such as summarizing text, sorting requests, or compressing long conversations to keep an AI agent's memory from being overloaded. It can also act as a sub-agent, assisting larger models like Opus 5.5 or Sonnet 5.5 in handling smaller tasks within a code project. While the price reduction is significant, it applies only to requests under 100,000 tokens. For larger requests, the cost increases to $0.50 for input and $2.50 for output, which is five times the base rate. Anthropic notes that 90% of the requests sent to Claude Haiku 4.5 fell under this threshold, making the new pricing particularly beneficial for many users. The new model also uses a different tokenizer, which slightly increases the number of tokens required for the same task. Despite this, Anthropic claims the overall cost is approximately 75% lower than with Claude Haiku 4.5. Anthropic has aligned its pricing with that of OpenAI's GPT-6 Luna, the smallest model from OpenAI, which also costs $0.10 for input and $0.50 for output. Anthropic also claims that Claude Haiku 5.5 outperforms GPT-6 Luna in several benchmarks, including OSWorld 2.1, where it achieved a score of 72.4% compared to 48.9% for Luna. However, Anthropic acknowledges that for complex tasks, larger models like Sonnet 5.5 and Opus 5.5 remain superior. While the new model shows promise, independent testing is still needed to confirm its performance.