4 min readNew DelhiOct 8, 2026 02:04 PM IST
It’s another day, another Anthropic AI model. The Frontier AI lab introduced yet another model in its Haiku class of AI models on Wednesday, October 8. Claude Haiku 5.5 is described as Anthropic’s cheapest, fastest, and most capable small model. It is designed to handle high-volume and cost-sensitive tasks.
Anthropic said that Haiku 5.5 can reliably manage quick and repetitive workloads such as summaries, compactions, database queries, and classification requests. When it comes to coding work, the model reportedly pairs well with Opus 5.5 and Sonnet 5.5. Also, Anthropic claims that the model works well for speed-sensitive tasks like live customer support and browser use. When it comes to pricing, Haiku 5.5 is available at a much lower price than Haiku 4.5. Anthropic claims that the model costs around 75 per cent less to run.
Haiku 5.5 is the third model in Anthropic’s Claude 5.5 family launched in a month, following Opus 5.5 and Sonnet 5.5. It is crucial for Anthropic as it steps up its AI model releases ahead of a potential IPO. Also, with Haiku 5.5 being positioned as a cheaper and faster model, Anthropic seems to not only expand its product lineup but also demonstrate its ability to cater to a wider range of customers and workloads amid growing investor scrutiny.
How does the model perform?
According to Anthropic, Haiku 5.5 performs across a range of benchmarks. On GDPval-AA v2.1, which tests abilities in terms of knowledge work, Haiku 5.5 secured 1620, outdoing OpenAI’s GPT-6 Luna, which scored 1437. When it comes to computer use, the model secured 74 per cent in OSWorld 2.1 against GPT-6 Luna’s 48.9 per cent. The model secured 45.9 per cent in Humanity’s Last Exam, a benchmark to ascertain advanced reasoning skills. Besides, Haiku 5.5 secured scores above GPT-6 Luna in agentic coding and visual reasoning.
Anthropic said that it is the first Haiku-class model to feature an adjustable effort setting. This essentially means that users can decide whether to optimise for cost or intelligence. The AI lab also shared three key benchmarks at different effort settings. It also shared testimonials from customers who attested to the consistency in performance and cost improvements claimed by the company.
How safe is Claude Haiku 5.5?
The latest model from Anthropic’s stable is accompanied by some major improvements across all of its alignment evaluations relative to Haiku 4.5. The company claims that it found fewer instances of misaligned behaviour and a lower willingness to partake in misuse. When it comes to safeguards, the model’s cybersecurity measures are more restrictive when compared to Haiku 4.5. However, it is less restrictive than the safeguards applied to other recent models. The model has similar biology safeguards as Sonnet 5, Sonnet 5.5, and Opus 5. These allow research biology questions but limit access to requests that the company deems likely to cause harm.
When it comes to pricing, Claude Haiku 5.5 is priced lower than Anthropic’s other models, especially for prompts of up to 100,000 tokens, which account for around 90 per cent of requests made to the previous Haiku model. It costs $0.10 per million input tokens and $0.50 per million output tokens, compared with $1 and $5 for Haiku 4.5, and $2 and $10 for Sonnet 5.5.
Story continues below this ad
Claude Haiku 5.5 is available on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. Meanwhile, developers can get started with Claude-Haiku-5.5 on the Claude Platform.
