TL;DR
Get tech for your team delivered free — and shop member deals
- Fast, free delivery on millions of items
- Access to Prime Big Deal Days deals on October 6–7
- Prime Video, Amazon Music and more included
Anthropic has released Claude Haiku 5.5, a small model aimed at fast, high-volume tasks, and says it costs about 75% less to run on average than Haiku 4.5. The model is available across Anthropic’s platforms and major cloud providers; its benchmark results and customer feedback are company-reported, and independent comparisons are not included in the announcement.
Anthropic has released Claude Haiku 5.5, a small language model it says is its fastest to date and costs about 75% less to run on average than Haiku 4.5. The company is targeting high-volume, speed-sensitive work such as summarization, classification, customer support and browser use, and says the model is available now through its own platform and major cloud providers.
Anthropic describes Haiku 5.5 as a model for quick, repetitive and cost-sensitive workloads, including summaries, context compaction, database queries and classification. It also says the model can serve as a subagent alongside its larger Opus 5.5 and Sonnet 5.5 models in coding work. Those are intended use cases, not independent findings about how the model performs in production.
The company’s published pricing lists Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens for prompts up to 100,000 tokens. For prompts over that threshold, the listed rates are $0.50 for input and $2.50 for output per million tokens. Anthropic says roughly 90% of requests to its previous Haiku model were within the lower prompt-length category. Its comparison table lists Haiku 4.5 at $1 per million input tokens and $5 per million output tokens.
Anthropic also announced a 50% reduction in Sonnet 5.5 cache-read pricing, from $0.20 to $0.10 per million tokens. It estimates that the change makes Sonnet 5.5 around 20% cheaper on most agentic tasks. The company said it is adding a monthly API credit for Claude Max and Team subscribers to support development on its platform; the announcement did not specify the credit amount.
Lower Costs for High-Volume AI Work
The release gives developers another option for workloads where response speed and per-request cost may matter more than top performance on complex tasks. If Anthropic’s pricing and performance claims hold for a particular application, teams could route routine requests to Haiku 5.5 while using larger models for harder work. That kind of division can affect the cost and responsiveness of products built around AI agents.
Anthropic’s own benchmark table also signals a limit to that proposition: its larger models score higher on some demanding tasks. For example, the company reports 39.2% on Terminal-Bench 4.0 for Haiku 5.5, compared with 70.6% for Sonnet 5.5. The figures come from Anthropic’s evaluations and should not be treated as independent verification. The practical choice will depend on each application’s task mix, latency needs, and actual costs.
As an affiliate, we earn on qualifying purchases.
How Haiku Fits Anthropic’s Model Range
Haiku is Anthropic’s smaller-model line, positioned for faster or less expensive work than its larger models. The company says Haiku 5.5 is its first Haiku model with an adjustable effort setting, allowing users to choose between lower cost and greater reasoning effort. Anthropic’s announcement includes results at different effort settings for selected benchmarks, but the supplied release does not provide a complete independent evaluation.
The model is available through the Claude Platform under the identifier claude-haiku-5-5, as well as through Amazon Web Services, Google Cloud and Microsoft Azure, according to Anthropic. The company points developers to a migration guide and a system card for details on evaluations and safeguards.
Anthropic’s release presents the new model as part of a broader pricing update, not only as a model launch. Along with the Haiku 5.5 rates, it reduced Sonnet 5.5 cache-read costs and announced API credits for Claude Max and Team subscribers. The announcement’s available details do not state the credit’s value or duration.
““the cheapest, fastest, and most capable small model we’ve ever released””
— Anthropic
As an affiliate, we earn on qualifying purchases.
Independent Results Still Pending
The announcement does not include independent benchmark results or enough detail to establish how Haiku 5.5 will perform across different customer workloads. Benchmark scores can vary with evaluation setup, model settings and task selection. Anthropic directs readers to its system card for methodology and additional results, but the company’s own figures remain vendor-reported.
Several commercial details are also unspecified in the release, including the amount of the monthly API credit for eligible Max and Team subscribers. Asana’s reported latency gains are based on its internal evaluation and do not establish that other customers will see the same improvement. The release does not provide independent confirmation of the average 75% cost reduction across all workloads.
Anthropic says Haiku 5.5’s cybersecurity safeguards are more restrictive than Haiku 4.5’s, while allowing a wider range of defensive tasks than Sonnet 5.5. It says the safeguards still block penetration testing and other techniques it considers more likely to be misused. The release summarizes the policy but does not detail how those restrictions will behave in every use case.
cost-effective AI summarization tool
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Availability and Developer Testing
Developers can begin testing Haiku 5.5 now through the Claude Platform and listed cloud providers, using the model identifier claude-haiku-5-5. Anthropic points users to its migration guide for implementation details and its system card for evaluation and safety information. Comparing the model with existing systems on representative tasks will help customers judge whether lower listed prices translate into savings for their particular workloads.
Anthropic has not provided a date for further model updates or stated when it will publish additional information about the API credit. For now, the next useful evidence will come from customer testing and evaluation details in the system card, alongside clearer terms for the subscriber credits.
high-volume AI classification software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Claude Haiku 5.5?
Claude Haiku 5.5 is Anthropic’s new small model, designed for fast, high-volume tasks such as summarization, classification, customer support and browser use.
How much does Haiku 5.5 cost?
Anthropic lists rates for prompts up to 100,000 tokens of $0.10 per million input tokens and $0.50 per million output tokens. Rates are higher for prompts over 100,000 tokens.
Where can developers use the model?
Anthropic says Haiku 5.5 is available through the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure. Its Claude Platform model identifier is claude-haiku-5-5.
Is Haiku 5.5 better than Sonnet 5.5 for coding?
Not across all coding tasks, based on Anthropic’s own benchmark results. The company says Sonnet 5.5 and Opus 5.5 remain better suited to complex agentic coding, while Haiku 5.5 is aimed at more narrowly scoped tasks. Its published scores have not been independently verified in the announcement.
What other pricing changes did Anthropic announce?
Anthropic cut Sonnet 5.5 cache-read pricing by 50%, to $0.10 per million tokens, and said that makes the model around 20% cheaper on most agentic tasks. It also announced monthly API credits for Claude Max and Team subscribers but did not specify the credit amount.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
