Anthropic launches Claude Haiku 5.5 with lower prices for high-volume tasks
Anthropic says its new small model costs about 75% less to run on average than Haiku 4.5 and adds adjustable effort and tighter limits on some cybersecurity requests.
Anthropic launched Claude Haiku 5.5 on October 7, 2026, offering developers a lower-priced model for frequent, smaller tasks such as summaries and classification. The company says it is available through its Claude Platform and major cloud providers. The release is the third Claude 5.5 model upgrade reported in about a month, following Opus and Sonnet.
The launch matters to teams that run large numbers of AI requests or need quick responses: Anthropic says Haiku 5.5 costs about 75% less to run on average than Haiku 4.5. That is the company's estimate, which accounts for changes in tokenization and the number of tokens used per task; an individual customer's savings will depend on its workload.
How Claude Haiku 5.5 pricing works
Anthropic lists prices of $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. For longer prompts, the listed prices rise to $0.50 and $2.50 respectively. The company lists Haiku 4.5 at $1 per million input tokens and $5 per million output tokens. Input tokens represent material sent to the model; output tokens are what it generates.
Anthropic describes the price reduction as 90% for requests up to 100,000 tokens and 50% for longer requests, while its 75% figure is an average cost-to-run estimate. It says roughly 90% of requests to the previous Haiku model used prompts within the lower-priced band. Those figures help explain the potential benefit for high-volume users, but the published averages do not set a fixed saving for every application.
Alongside the Haiku release, Anthropic says it halved Sonnet 5.5 cache-read prices, from $0.20 to $0.10 per million tokens. It estimates that change makes Sonnet 5.5 around 20% cheaper for most agentic work. The company also says monthly Claude Platform API credits will begin rolling out during the week of the announcement: $100 for Max 5x subscribers, $200 for Max 20x subscribers and up to $500 pooled across Team subscribers.
Availability and intended uses
Anthropic says Haiku 5.5 is available now on its Claude Platform under the model ID claude-haiku-5-5, as well as through Amazon Web Services, Google Cloud and Microsoft Azure. That availability is the company's account; the cited reporting does not separately verify deployment on each cloud platform.
The company pitches the model for high-volume, cost-sensitive work including summaries, database queries and classification. It also names live customer support and browser use as tasks where response speed matters. Haiku 5.5 is the first Haiku-class model with an adjustable effort setting, according to Anthropic, allowing users to choose a setting that favors lower cost or more intensive reasoning.
Anthropic presents Haiku as an option for narrower work, including subagent tasks within coding systems. It says Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding tasks of the kind measured by Terminal-Bench 4.0. That distinction matters for developers comparing the cheaper model with the rest of the Claude range: the launch does not mean Anthropic is positioning Haiku as its strongest model for every job.
What the performance and safety claims establish
In Anthropic's published evaluations, Haiku 5.5 scored 39.2% on Terminal-Bench 4.0, compared with 0.0% for Haiku 4.5. It reported 72.4% on the offline subset of OSWorld 2.1 against 15.7% for its predecessor, and 45.9% on Humanity's Last Exam without tools against 10.2%. These are company-published benchmark results, not a measure of how the model will perform in every production setting.
Anthropic also includes an early customer account from Asana. Aaron Vinh, a staff software engineer, said an evaluation for Asana's AI Teammates product found more than 30% lower latency for task completions and up to 2.5 times faster inference per agent turn than the model Asana currently uses. The account appears as a customer testimonial on Anthropic's launch page; it does not establish the same gain for other users or tasks.
On safety, Anthropic says Haiku 5.5 applies more restrictive cybersecurity safeguards than Haiku 4.5. It says the model blocks some penetration-testing requests and techniques it considers more likely to be used by attackers, although the limits are less restrictive than those on some of its other recent models. Anthropic says its biology safeguards still allow research questions while restricting requests it judges likely to cause harm. The practical effect of those boundaries for individual users is not established by the launch announcement.
Sources and context
AI-assisted article checked against the listed sources. NewsJaws did not conduct interviews or attend the reported events.
About NewsJaws Desk
AI-assisted reporting and explainers reviewed against the linked source documents. No claim of on-scene reporting or original interviews.