News in Short
- Anthropic has launched Claude Haiku 5.5, its fastest and most capable small model to date.
- The company says Haiku 5.5 costs around 75% less to run than Haiku 4.5 on average.
- The model targets repetitive tasks, coding subagents, browser use and live customer support.
- Anthropic has cut Claude Sonnet 5.5 cache-read prices by 50%, reducing costs for many agentic tasks.
- Claude Max and Team subscribers will receive monthly API credits to build applications and AI agents.
Anthropic has launched Claude Haiku 5.5, a small AI model designed to deliver faster performance at a lower operating cost. The company is targeting developers and businesses that run AI agents across large volumes of everyday tasks.
Haiku 5.5 handles workloads such as summarisation, classification, database queries and customer support. It can also work alongside Anthropic’s larger models on coding tasks, acting as a subagent for more complex workflows.
Anthropic says Haiku 5.5 costs around 75% less to run than its predecessor, Haiku 4.5. Alongside the launch, the company has reduced cache-read pricing for Claude Sonnet 5.5 and introduced monthly API credits for eligible subscribers.
Claude Haiku 5.5 focuses on speed and high-volume AI tasks
Anthropic designed Haiku 5.5 for workloads where speed and cost matter as much as capability. These include repetitive tasks that businesses run at scale, such as summarising documents, classifying information and compressing long conversations.
The model is also suited to live customer support and browser-based tasks. Its speed could help reduce delays when AI agents need to interact with websites or complete multiple steps.
Developers can also pair Haiku 5.5 with Claude Sonnet 5.5 and Opus 5.5. In these setups, Haiku can handle smaller subtasks while a larger model tackles more complex reasoning or coding work.
Anthropic says its customers have already seen improvements in early testing. Asana reported more than a 30% reduction in task-completion latency and up to 2.5 times faster inference per agent turn for its AI teammate product.
Haiku 5.5 posts gains across several benchmarks
Anthropic’s published benchmark results show significant improvements over Haiku 4.5 across knowledge work, computer use, reasoning and coding.
| Benchmark | Claude Haiku 5.5 | Claude Haiku 4.5 |
|---|---|---|
| GDPval-AA v2.1 | 1,620 | 735 |
| AA-Briefcase v1.1 | 1,578 | 614 |
| OSWorld 2.1 | 72.4% | 15.7% |
| Humanity’s Last Exam, no tools | 45.9% | 10.2% |
| Terminal-Bench 4.0 | 39.2% | 0.0% |
| FrontierCode 1.1 | 46.4% | Not listed |
These figures come from Anthropic’s own evaluations. They indicate progress across several task categories, although benchmark results do not guarantee similar performance in every real-world application.
Haiku 5.5 also introduces adjustable effort settings for the first time in the Haiku family. Users can choose between lower-cost responses and greater reasoning effort, depending on the task.
However, it is not intended to replace Anthropic’s larger models in every situation. Sonnet 5.5 and Opus 5.5 remain better suited to demanding agentic coding tasks, according to the company’s results.
Anthropic cuts API pricing for Haiku 5.5
The biggest commercial change is the reduction in token costs. Anthropic says Haiku 5.5 costs around 75% less to run than Haiku 4.5 on average.
The pricing makes Haiku 5.5 particularly attractive for applications that process large numbers of relatively short prompts. Anthropic says around 90% of requests to its previous Haiku model fell within the 100,000-token threshold.
For businesses running AI agents continuously, lower token costs could make previously expensive workflows more practical.
Claude Sonnet 5.5 also becomes cheaper
Anthropic is reducing costs for its larger model as well. It has halved Claude Sonnet 5.5’s cache-read price from $0.20 to $0.10 per million tokens.
The company estimates this change will make Sonnet 5.5 around 20% cheaper for most agentic workloads. The impact will depend on how much cached input a particular application uses.
This move is significant because AI agents often reuse instructions and context across multiple steps. Lower cache-read costs can therefore reduce the expense of running longer workflows without requiring developers to switch to a smaller model.
Monthly API credits and wider availability
Anthropic is also introducing monthly API credits for Claude Max and Team subscribers. The credits will support experimentation with applications, tools and agents built on the Claude Platform.
- Max 5x subscribers will receive $100 in monthly API credits.
- Max 20x subscribers will receive $200 in monthly API credits.
- Team subscribers will receive up to $500 in monthly credits, pooled across users.
The credits can be used across Anthropic’s models. The company is also adding beta support for computer use and browser use to its Python and TypeScript SDKs, opening up more ways for developers to build automated workflows.
Claude Haiku 5.5 is available through the Claude Platform and cloud platforms including Amazon Web Services, Google Cloud and Microsoft Azure.
With this launch, Anthropic is making a broader push to reduce the cost of deploying AI agents. Haiku 5.5 targets frequent, narrowly defined tasks, while lower Sonnet pricing makes larger-model workflows more economical. For businesses scaling AI beyond occasional chatbot queries, those cost reductions could matter as much as raw benchmark performance.
