Anthropic Launches Claude Haiku 5.5 Ahead of IPO
Anthropic has launched Claude Haiku 5.5, completing its three-model Claude 5.5 lineup. The release comes as the AI company expands its product range ahead of a planned initial public offering.
Haiku 5.5 is designed for high-volume, cost-sensitive workloads. In particular, Anthropic targets summarisation, classification, database queries, live customer support, voice agents and browser-based tasks.
The launch follows Claude Opus 5.5 on September 22 and Claude Sonnet 5.5 on September 28. Therefore, Anthropic has introduced all three Claude 5.5 models within roughly two weeks.
Haiku 5.5 Targets High-Volume AI Work
Anthropic describes Haiku 5.5 as its fastest, cheapest and most capable small model to date. Moreover, the company says it costs around 75% less to run than Haiku 4.5 on average.
For prompts up to 100,000 tokens, Haiku 5.5 costs $0.10 per million input tokens. Meanwhile, output costs $0.50 per million tokens. For longer prompts, the rates rise to $0.50 for input and $2.50 for output.
The pricing represents a major reduction from Haiku 4.5. Previously, that model cost $1 per million input tokens and $5 per million output tokens. As a result, the new short-prompt rates are 90% lower per token.
However, Anthropic’s 75% average cost reduction reflects actual workload patterns rather than the headline token-rate cut. The company says that about 90% of requests to its previous Haiku model were under 100,000 tokens.
Haiku 5.5 also introduces adjustable effort, a first for Anthropic’s Haiku family. Consequently, developers can adjust how much reasoning effort the model applies to individual tasks.
The model can also work alongside larger Claude models as a subagent. For example, developers can use Haiku for repetitive tasks while assigning complex reasoning to Sonnet or Opus.
Anthropic Adds New Safety Controls
Alongside its performance and pricing changes, Anthropic has introduced additional safety measures. Haiku 5.5 includes more restrictive cybersecurity safeguards than Haiku 4.5.
However, its protections remain somewhat less restrictive than those used with Anthropic’s other recent models. The company says the model supports a broader range of defensive cybersecurity tasks while blocking penetration testing and other activities that could assist attackers.
Its biology safeguards also align with those used for Sonnet 5, Sonnet 5.5 and Opus 5. Anthropic says these safeguards allow legitimate research questions while restricting requests it considers likely to cause harm.
Furthermore, Anthropic reports significant improvements across its alignment evaluations. The company says Haiku 5.5 showed fewer instances of misaligned behaviour and less willingness to cooperate with misuse.
Haiku 5.5 is available across Anthropic’s major cloud platforms. Developers can access it through the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure.
Third Model Strengthens Anthropic’s AI Portfolio
The new release gives Anthropic a complete three-tier Claude 5.5 lineup. Opus 5.5 targets complex reasoning and demanding agentic work. Sonnet 5.5 focuses on general-purpose tasks, coding and professional workflows. Meanwhile, Haiku 5.5 handles speed-sensitive and high-volume workloads.
Sonnet 5.5 is more than 30% faster than its predecessor and can cost up to 30% less per task, according to Anthropic. It also targets everyday coding, document creation and other well-defined workloads.
At the top of the range, Opus 5.5 serves more demanding work requiring sustained judgment. Therefore, the three-model structure gives Anthropic a clearer way to match computing costs with workload requirements.
The timing also matters as Anthropic prepares for a potential public listing. In June, the company confirmed that it had confidentially submitted a draft S-1 registration statement to the U.S. Securities and Exchange Commission. Anthropic said the proposed IPO would depend on market conditions and other factors.
Reuters reported that Anthropic’s latest model rollout forms part of its efforts to expand its AI portfolio ahead of the planned IPO. The company has therefore entered the public-market process while continuing to compete with OpenAI, Google and other major AI developers.
Ultimately, Haiku 5.5 gives Anthropic a lower-cost option for deploying AI at scale. At the same time, its three-tier Claude 5.5 family lets customers choose between deeper reasoning, balanced performance and high-volume efficiency. That combination could become increasingly important as enterprises move more AI workloads into production.
