Anthropic Launches Claude Haiku 5.5: From $0.10, About 75% Cheaper Than Its Predecessor

What Happened

The numbers first: on October 7 (ET), Anthropic launched Claude Haiku 5.5 — the third and final model of the Claude 5.5 family after Opus 5.5 (September 22) and Sonnet 5.5 (September 28). Positioned as the fastest and most efficient model in the family, built for high-volume, cost-sensitive work. Pricing is split by prompt length: $0.10 input / $0.50 output per million tokens for prompts up to 100K tokens; $0.50 / $2.50 beyond 100K. Anthropic says workloads run about 75% cheaper than Haiku 4.5 (per Reuters); VentureBeat puts the nominal price cut at ~90% — the two figures measure different things (see Key Facts k2).

Availability: Claude.ai (Free / Pro / Max / Team / Enterprise, web, iOS, Android), the Claude Platform API (model id claude-haiku-5-5), AWS / Google Cloud / Microsoft Foundry, and Claude Code. Three notable firsts: (1) the first Haiku with effort controls — teams can trade money for intelligence per task; (2) the first Haiku with built-in safeguards for a narrow set of high-risk cybersecurity requests (Anthropic says everyday tasks are unaffected); (3) Anthropic calls it the most capable Haiku ever — a step up over Haiku 4.5 across coding, tool use, computer use, and agents. Customer evidence: Box scored 11 points higher at about half the latency; Rogo ran 400 queries at 0.84 vs 0.76; HubSpot hit 92.8% on its CRM eval suite (the best score it has ever seen); Devin Fusion with Haiku 5.5 as sidekick scored 66.2 on FrontierCode.

Key Facts

1. $0.10 / $0.50 applies only to prompts up to 100K tokens: The $0.10 / $0.50 tier covers only prompts up to 100K tokens; beyond that it is $0.50 / $2.50, half of Haiku 4.5's $1 / $5. VentureBeat notes that this pricing matches OpenAI's GPT-6 Luna.
2. 75% vs 90%: two different measures: Reuters' "75% cheaper" is Anthropic's blended figure — total workload cost accounting for request-size distribution and changes in token consumption. VentureBeat's "90% price cut" is the nominal unit-price comparison ($1 down to $0.10).
3. Haiku's role in the 5.5 family: Anthropic positions Haiku as the fastest and most efficient model in the family, for high-volume, cost-sensitive work. Rogo's usage: the big model builds the earnings deck while Haiku extracts the revenue line from the 10-K.
4. The third Claude 5.5 model in a month: Reuters reports this is Anthropic's third Claude 5.5 model in a month, "expanding its AI lineup ahead of a planned IPO." Opus 5.5 on September 22, Sonnet 5.5 on September 28, Haiku 5.5 on October 7.

Context

Haiku 4.5 launched October 15, 2025, billed then as the "fastest, most cost-efficient model," scoring 73.3% on SWE-bench Verified. Anthropic set Haiku 4.5's retirement at "not sooner than October 15, 2026" — about one week after Haiku 5.5's launch. Haiku 5.5 got no standalone launch event — just a model page and API availability, the same release approach as Opus 5.5 and Sonnet 5.5.