Anthropic launched Claude Haiku 5.5 on October 7, 2026 — its third Claude 5.5 model in a month, expanding its AI lineup ahead of a planned IPO. Haiku is Anthropic's fastest, cheapest model family, and this release is aimed squarely at high-volume production workloads: classification, summarization, extraction, live customer support, and voice agents, according to Reuters.

The headline number is cost: Haiku 5.5 costs 75% less than Haiku 4.5, priced at $0.10/$0.50 per million input/output tokens for prompts under 100k tokens — rising to $0.50/$2.50 above 100k tokens. That aggressive pricing puts Anthropic in direct competition with Google and OpenAI's cheapest tiers at a moment when developers running millions of calls a day are extremely price-sensitive. For voice agents and live support bots, where every response must be near-instant and cheap, a 75% cost cut can change the economics of an entire product.

Notably, Haiku 5.5 is the first Haiku model with built-in safeguards for a narrow set of high-risk cybersecurity requests. That is unusual for a small, fast model — typically safety work focuses on flagship models — and it signals that Anthropic expects Haiku 5.5 to be deployed at real scale, where even edge-case misuse matters. The company has been expanding the Claude 5.5 lineup rapidly: recent weeks saw the launches of Sonnet 5.5 and Opus 5.5, making Haiku 5.5 the final piece of the refresh.

The timing is strategic. With Anthropic reportedly preparing for an IPO, a complete and competitive model lineup across all three tiers — Haiku, Sonnet, Opus — tells investors the company can fight Google's Gemini and OpenAI's GPT-6 on both quality and price. The real test will be adoption: whether developers running production workloads at scale switch their cheapest, highest-volume traffic to Haiku 5.5 now that it is this much cheaper to run.

Sources