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, with prices substantially below Haiku 4.5. The company reports benchmark and safety improvements, but those results come from Anthropic’s evaluations; independent comparisons and real-world performance remain to be established.
Anthropic has released Claude Haiku 5.5, a small model it says is designed for fast, high-volume workloads and costs around 75% less to run on average than Haiku 4.5. The launch also brings a 50% cut to Sonnet 5.5 cache-read prices and a new monthly API credit for Claude Max and Team subscribers.
Anthropic recommends Haiku 5.5 for repetitive or speed-sensitive tasks such as summarization, classification, database queries, live customer support and browser use. It also says the model can act as a subagent alongside Sonnet 5.5 or Opus 5.5 on coding work. The company describes Haiku 5.5 as its fastest model to date; that characterization is Anthropic’s, rather than a finding from an independent test.
For prompts up to 100,000 tokens, Anthropic lists Haiku 5.5 at $0.10 per million input tokens and $0.50 per million output tokens. For prompts over 100,000 tokens, those rates rise to $0.50 and $2.50, respectively. The listed rates for Haiku 4.5 are $1 per million input tokens and $5 per million output tokens. Anthropic says prompts up to 100,000 tokens made up around 90% of requests to its previous Haiku model.
The release is available on the Claude Platform and through Amazon Web Services, Google Cloud and Microsoft Azure, according to Anthropic. Developers can access it using the model identifier claude-haiku-5-5. Anthropic also says Haiku 5.5 is its first Haiku-class model with an adjustable effort setting, letting users trade off cost and capability.
Lower Costs for Routine AI Work
The pricing and speed focus could make it less expensive for companies to use AI across large volumes of routine requests, rather than reserving a more costly model for every step. Anthropic’s intended division of work is that Haiku handles narrow, frequent tasks, while larger models take on harder work. That may appeal to developers building agents, where repeated calls can add up in both latency and token costs.
Anthropic also cut Sonnet 5.5 cache-read pricing by 50%, from $0.20 to $0.10 per million tokens. The company estimates this will make Sonnet 5.5 around 20% cheaper on most agentic work, though the savings will depend on how a particular application uses cached inputs. The new monthly API credit for Claude Max and Team subscribers is intended to support building applications on the Claude Platform; the supplied announcement does not give the credit amount or detailed eligibility terms.
These changes matter to organizations comparing the cost of deploying AI products, but price alone does not establish overall value. Buyers still need to test whether Haiku 5.5 meets their accuracy, reliability, latency and safety requirements on their own tasks.
As an affiliate, we earn on qualifying purchases.
How Haiku Fits Anthropic’s Lineup
Anthropic presents Haiku 5.5 as a model for narrower, cost-sensitive tasks, not a replacement for its larger models on demanding work. Its release materials say Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding tasks, using Terminal-Bench 4.0 as an example. That distinction is relevant for developers deciding whether to assign a task to a smaller model or pay for a larger one.
Anthropic’s published benchmark table compares Haiku 5.5 with Haiku 4.5, Sonnet 5.5 and other listed models across knowledge work, computer use, reasoning, coding and visual reasoning. For example, the table reports a 39.2% score for Haiku 5.5 on Terminal-Bench 4.0, compared with 0.0% for Haiku 4.5 and 70.6% for Sonnet 5.5. These are company-published evaluation results, and the figures apply to the named benchmark rather than all coding work.
On safety, Anthropic says Haiku 5.5 improved on Haiku 4.5 in almost all of its alignment evaluations, including fewer instances of behavior it classified as misaligned and less willingness to cooperate with misuse. The company says its cybersecurity safeguards allow a wider range of defensive tasks than Sonnet 5.5’s safeguards, while still blocking penetration testing and other techniques it considers more likely to be used by attackers.
““Claude Haiku 5.5 is designed for high-volume, cost-sensitive tasks.””
— Anthropic
As an affiliate, we earn on qualifying purchases.
Independent Results Still Pending
The announcement’s performance figures come from Anthropic’s benchmark evaluations, and the material does not provide independent replication of those results. The company points readers to a system card for evaluation details, but the supplied information does not include the full testing methodology, uncertainty ranges or results from independent researchers.
Real-world cost and latency will vary with prompt length, output size, cache use, task design and how often a model must retry or hand work to another system. Asana’s reported gains reflect its own evaluation suite and comparison model; the announcement does not name that model or provide enough detail to generalize the results across customers.
The announcement also does not specify the amount or full terms of the monthly API credit for Claude Max and Team subscribers. Anthropic describes its safety findings and safeguards, but the supplied material does not establish how the model will perform across every deployment or type of misuse.
large language model for customer support
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Testing in Customer Workflows
Developers can begin evaluating Haiku 5.5 through the Claude Platform and the listed cloud providers. Anthropic directs customers to a migration guide and its system card for implementation and evaluation information. The next useful evidence will come from customer tests that compare the model with existing systems on representative tasks, including accuracy, response time, total cost and error rates.
Anthropic has not specified when it will publish further results or detailed terms for the subscriber API credit in the supplied announcement. Readers assessing the launch should watch for those details, along with independent evaluations and reports from organizations using the model in production.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is Claude Haiku 5.5?
It is a small AI model released by Anthropic, intended for fast, high-volume and cost-sensitive work such as summarization, classification and database queries.
How much does Haiku 5.5 cost?
Anthropic lists input at $0.10 per million tokens and output at $0.50 per million tokens for prompts up to 100,000 tokens. For longer prompts, the listed prices are $0.50 per million input tokens and $2.50 per million output tokens.
Where can developers access the model?
Anthropic says Haiku 5.5 is available on 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 Anthropic’s larger models?
Not for every task. Anthropic positions it for speed and lower-cost, narrower work, and says Sonnet 5.5 and Opus 5.5 remain better choices for complex agentic coding tasks.
Are the reported performance gains independently verified?
The supplied launch material reports Anthropic’s own benchmark results and a customer’s early testing experience. It does not provide independent confirmation of those findings.
Source: hn
Halloween Picks
halloween
As an affiliate, we earn on qualifying purchases.
