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 such as summarization, classification and customer support. The company says it costs about 75% less to run on average than Haiku 4.5 and is available across its platform and major cloud services. Benchmark results and customer performance figures cited in the announcement are Anthropic’s and its customers’ claims.
Anthropic has released Claude Haiku 5.5, a small language model designed for quick, high-volume work, and says it costs about 75% less to run on average than its predecessor, Haiku 4.5. The launch also brings a lower cache-read price for Sonnet 5.5 and a new monthly API credit for Claude Max and Team subscribers.
Anthropic positions Haiku 5.5 for workloads where speed and cost matter, including summarization, prompt compaction, database queries, classification, live customer support and browser use, considerations explored in switching Claude models. The company says it is its fastest model to date and recommends pairing it with Sonnet 5.5 or Opus 5.5 as a subagent for coding work. It also describes Haiku as better suited to narrower tasks than complex, multi-step agentic coding.
The new model is available on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure, while Microsoft’s Claude spending has also drawn attention. Anthropic’s published prices per million tokens are $0.10 for input and $0.50 for output for prompts up to 100,000 tokens. For prompts above that length, the listed rates are $0.50 for input and $2.50 for output. Anthropic says around 90% of requests to its previous Haiku model were within the lower prompt-length tier, a detail relevant to comparisons of Claude’s value.
Anthropic also says Haiku 5.5 is the first Haiku-class model with an adjustable effort setting, letting users choose between cost and performance. Its announcement reports improvements over Haiku 4.5 on most of its alignment evaluations, while describing the model’s cybersecurity safeguards as more restrictive than Haiku 4.5’s. The company says those safeguards block penetration testing and other techniques it considers more likely to be used by attackers.
Lower Costs for Frequent Model Calls
For developers running agents or other software that makes repeated model calls, per-token pricing and response speed can shape operating costs as much as a model’s performance. Anthropic’s stated average 75% reduction compared with Haiku 4.5 could make routine tasks more economical, though the actual savings will depend on prompt size, output volume and usage patterns.
The launch also adds a cheaper option to a model lineup in which Anthropic says its larger systems remain preferable for harder coding work. That division gives businesses a possible way to use Haiku for bounded, repetitive steps and reserve more capable models for tasks that need them. The company’s own benchmarks and early customer feedback offer an initial picture, not independent evidence of performance across all deployments.
The accompanying Sonnet 5.5 cache-read price cut broadens the announcement beyond Haiku. Anthropic says the reduction makes Sonnet about 20% cheaper on most agentic work, a company estimate that will vary by workload. Monthly API credits for Claude Max and Team subscribers may also help those users experiment with applications on the Claude Platform; the announcement does not give the credit amount.
AI language model API for developers
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
How Haiku Fits Anthropic’s Lineup
Anthropic’s announcement presents Haiku 5.5 as a model for high-volume, cost-sensitive tasks, rather than as a replacement for its larger models in every use case. The company specifically says Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding tasks like those measured by Terminal-Bench 4.0.
In its benchmark table, Anthropic reports Haiku 5.5 scores of 39.2% on Terminal-Bench 4.0 and 46.4% on FrontierCode 1.1 (Main). Sonnet 5.5 is listed at 70.6% and 52.1%, respectively. These results are vendor-reported benchmark figures; the announcement points readers to the model system card for details on evaluation methods. The comparisons should be read alongside the benchmarks’ specific tasks and conditions, not as a single measure of general ability.
Haiku 5.5’s listed prices are lower than Haiku 4.5’s: Anthropic gives Haiku 4.5 rates of $1 per million input tokens and $5 per million output tokens. For Sonnet 5.5, the company separately cuts cache reads by half, to $0.10 per million tokens, and says this lowers costs on most agentic tasks by around 20%.
“Claude Haiku 5.5 is the cheapest, fastest, and most capable small model we’ve ever released.”
— Anthropic
Performance Beyond Anthropic’s Tests
The announcement does not specify the date of release, and it does not provide the amount or eligibility details for the new monthly API credit. Anthropic’s statement that Haiku 5.5 costs about 75% less to run on average is not tied in the supplied material to a full breakdown of workloads, token use or the calculation method.
Benchmark scores and safety evaluation findings are reported by Anthropic, while the latency and inference-speed figures come from an early customer evaluation at Asana. The source does not provide independent replication or results across a wider range of users. Real-world costs and performance remain workload-dependent, and the announcement does not establish how the model compares across every task or deployment environment.
Testing Costs on Real Workloads
Developers can begin using Haiku 5.5 through the Claude Platform and the named cloud providers; Anthropic directs users to a migration guide for moving to the new model. The announcement does not give a later rollout milestone, so the next practical test for customers will be how the model performs on their own tasks, at their chosen effort setting and prompt lengths.
Users considering the Sonnet 5.5 price change will need to compare their cache-read usage and total API bills to see whether Anthropic’s estimated savings apply to their workloads. Further details about the monthly API credit—including its value and terms—are also still needed to assess how much support it provides for building applications.
Key Questions
What is Claude Haiku 5.5?
Claude Haiku 5.5 is Anthropic’s small language model for fast, high-volume tasks such as summaries, classification, database queries, customer support and browser use.
How much does Haiku 5.5 cost?
Anthropic lists prices per million tokens of $0.10 for input and $0.50 for output for prompts up to 100,000 tokens. For longer prompts, it lists $0.50 for input and $2.50 for output.
Where can developers use the model?
Anthropic says Haiku 5.5 is available on the Claude Platform, Amazon Web Services, Google Cloud and Microsoft Azure. Its model identifier on the Claude Platform is claude-haiku-5-5.
Is Haiku 5.5 intended for complex coding tasks?
Anthropic recommends it as a subagent for coding work but says Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding tasks. Haiku is aimed more at narrower, repetitive tasks where speed and cost are priorities.
What else changed in Anthropic’s pricing?
Anthropic cut Sonnet 5.5 cache-read pricing by 50%, to $0.10 per million tokens. The company estimates this makes Sonnet about 20% cheaper on most agentic work, though savings depend on usage.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
